Subclass 9	Subclass 10	Subclass 100
Distributed or remote access

System and method for determining relevancy of query responses in a distributed network search mechanism

6961723

Abstract

A system and method for selecting or ordering search results received from members of a distributing search network in response to a search request. Network nodes operating as consumer or requesting nodes generate the search requests. Nodes operating as hubs are configured to route the search requests in the network. Individual nodes operating as provider nodes receive the search request and in response may generate results according to their own procedures and return them. Communication between nodes in the network may use a common query protocol. Hub nodes may resolve the search requests to a subset of the provider nodes in the network, for example by matching search requests with registration information from nodes. Search results may be selected, ordered, and/or consolidated for use by the requesting nodes by nodes receiving a plurality of the search results.

Claims

1. A method for searching distributed resources, comprising:

receiving a search response formatted in accordance with a common query protocol from each of a plurality of provider nodes in a network in response to a search query formatted in accordance with the common query protocol from a requesting node in the network, wherein the search response includes relevance information; and

generating an aggregate search result formatted in accordance with the common query protocol including one or more of the plurality of search responses ordered in response to relevance information indicating a ranking of the plurality of search responses.

2. The method as recited in claim 1, wherein the relevance information indicates a ranking of each of the plurality of search responses according to the corresponding provider node.

3. The method as recited in claim 1, further comprising, prior to said generating an aggregate search result, generating additional relevance information in response to the relevancy of preceding search responses of at least one of the plurality of requesting nodes.

4. The method as recited in claim 1, further comprising, prior to said generating an aggregate search result, generating additional relevance information in response to user data from the corresponding requesting node.

5. The method as recited in claim 1, wherein each of the one or more search responses matches at least one of additional search parameters generated in response to the plurality of search responses being more than a specified number.

6. A method comprising:

sending a search query to first and second provider nodes in a network;

receiving first and second search responses from the first and second provider nodes respectively in response to the search query, wherein the first and second search responses include first and second bid numbers respectively; and

generating a collated search result including the first and second search response ordered in response to the difference between the first and second bid numbers.

7. The method as recited in claim 6, further comprising allocating a first and second number of tokens to the first and second provider nodes respectively.

8. The method as recited in claim 7, wherein the first bid number is not greater than the first number of tokens and the second bid number is not greater than the second number of tokens.

9. The method as recited in claim 7, further comprising subtracting the first and second bid numbers from the first and second number of tokens respectively.

10. The method as recited in claim 7, further comprising adding a third number of tokens to the first number of tokens in response to user feedback data indicating the first search response.

11. The method as recited in claim 7, further comprising adding a third number of tokens to the first and second number of tokens at certain time intervals.

12. A computer system in a network, comprising:

a network interface configured to receive a search response formatted in accordance with a common query protocol from each of a plurality of provider nodes in the network in response to a search query formatted in accordance with the common query protocol from a requesting node in the network and to transmit an aggregate search response to the requesting node, wherein the search response includes relevance information;

a storage device configured to store the plurality of search responses; and

a processing unit configured to generate an aggregate search result formatted in accordance with the common query protocol including one or more of the plurality of search responses ordered in response to relevance information indicating a ranking of the plurality of search responses.

13. The computer system as recited in claim 12, wherein the relevance information indicates a ranking of each of the plurality of search responses according to the corresponding provider node.

14. The computer system as recited in claim 12, wherein the processing unit is configured to generate additional relevance information in response to the relevancy of preceding search responses of at least one of the plurality of requesting nodes prior to said generating an aggregate search result.

15. The computer system as recited in claim 12, wherein the processing unit is configured to generate additional relevance information in response to user data from the corresponding requesting node prior to said generating an aggregate search result.

16. The computer system as recited in claim 12, wherein each of the one or more search responses matches at least one of additional search parameters generated in response to the plurality of search responses being more than a specified number.

17. The computer system as recited in claim 12, wherein the one or more search responses are selected from the plurality of search responses in response to the relevance information.

18. The computer system as recited in claim 12, wherein the one or more search responses are selected from the plurality of search responses in response to other relevance information indicating another ranking of the plurality of search responses.

19. A computer system in a network, comprising:

a network interface configured to send a search query from a requesting node in the network to first and second provider nodes in the network and to receive first and second search responses from the first and second provider nodes respectively in response to the search query, wherein the first and second search responses include first and second bid numbers respectively;

a storage device configured to store the first and second search responses; and

a processing unit configured to generating a collated search result including the first and second search response ordered in response to the difference between the first and second bid numbers.

20. The computer system of claim 19, wherein the network interface is configured to transmit the collated search result to the requesting node.

21. The computer system as recited in claim 19, wherein tokens are allocated to the first and second provider nodes and the first and second bid numbers do not exceed the respective allocated tokens.

22. The computer system as recited in claim 21, wherein the number of tokens allocated to the first and second provider nodes are reduced by the first and second bid numbers respectively.

23. The method as recited in claim 22, wherein additional tokens are allocated to the first and second nodes in response to specified events.

24. A computer system in a network, comprising:

means for receiving a search response formatted in accordance with a common query protocol from each of a plurality of provider nodes in the network in response to a search query formatted in accordance with the common query protocol from a requesting node in the network, wherein the search response includes relevance information;

means for storing the plurality of search responses;

means for transmitting an aggregate search response to the requesting node; and

means for generating an aggregate search result formatted in accordance with the common query protocol including one or more of the plurality of search responses ordered in response to relevance information indicating a ranking of the plurality of search responses.

25. The computer system as recited in claim 24, wherein the relevance information indicates a ranking of each of the plurality of search responses according to the corresponding provider node.

26. The computer system as recited in claim 24, further comprising means for generating additional relevance information in response to search responses received from at least one of the plurality of requesting nodes prior to said generating an aggregate search result for preceding search requests.

27. The computer system as recited in claim 24, further comprising means for generating additional relevance information in response to user data from the corresponding requesting node prior to said generating an aggregate search result.

28. The computer system as recited in claim 24, further comprising means for generating at least one additional search parameter in response to the plurality of search responses being more than a specified number, wherein each of the one or more search responses matches at least one of the at least one additional search parameter.

29. The computer system as recited in claim 24, further comprising means for selecting the one or more search responses from the plurality of search responses in response to the relevance information.

30. The computer system as recited in claim 24, further comprising means for selecting from the plurality of search responses the one or more search responses in response to other relevance information indicating another ranking of the plurality of search responses.

31. A computer system in a network, comprising:

means for transmitting a search query from a requesting node in the network to first and second provider nodes in the network;

means for receiving first and second search responses from the first and second provider nodes respectively in response to the search query, wherein the first and second search responses include first and second bid numbers respectively;

means for generating a collated search result including the first and second search response ordered in response to the difference between the first and second bid numbers.

32. The computer system of claim 31, further comprising means for transmitting the collated search result to the requesting node.

33. The computer system as recited in claim 31, further comprising means for allocating tokens to the first and second provider nodes, wherein the first and second bid numbers do not exceed the respective allocated tokens.

34. The computer system as recited in claim 33, further comprising means for reducing the allocated tokens to the first and second provider nodes by the first and second bid numbers respectively.

35. The computer system as recited in claim 34, further comprising means for allocated additional tokens to the first and second nodes in response to specified events.

36. A method for searching distributed resources, comprising:

receiving a search response formatted in accordance with a common query protocol from each of a plurality of provider nodes in a network in response to a search query formatted in accordance with the common query protocol from a requesting node in the network; and

generating an aggregate search result formatted in accordance with the common query protocol including one or more of the plurality of search responses ordered in response to relevance information indicating a ranking of the plurality of search responses;

wherein each of the one or more search responses matches at least one of additional search parameters generated in response to the plurality of search responses being more than a specified number.

37. A computer system in a network, comprising:

a network interface configured to receive a search response formatted in accordance with a common query protocol from each of a plurality of provider nodes in the network in response to a search query formatted in accordance with the common query protocol from a requesting node in the network and to transmit an aggregate search response to the requesting node;

a storage device configured to store the plurality of search responses;

a processing unit configured to generate an aggregate search result formatted in accordance with the common query protocol including one or more of the plurality of search responses ordered in response to relevance information indicating a ranking of the plurality of search responses; and

wherein each of the one or more search responses matches at least one of additional search parameters generated in response to the plurality of search responses being more than a specified number.

38. A computer system in a network, comprising:

a network interface configured to receive a search response formatted in accordance with a common query protocol from each of a plurality of provider nodes in the network in response to a search query formatted in accordance with the common query protocol from a requesting node in the network and to transmit an aggregate search response to the requesting node;

a storage device configured to store the plurality of search responses;

a processing unit configured to generate an aggregate search result formatted in accordance with the common query protocol including one or more of the plurality of search responses ordered in response to relevance information indicating a ranking of the plurality of search responses; and

wherein the one or more search responses are selected from the plurality of search responses in response to other relevance information indicating another ranking of the plurality of search responses.

39. A computer system in a network, comprising:

means for receiving a search response formatted in accordance with a common query protocol from each of a plurality of provider nodes in the network in response to a search query formatted in accordance with the common query protocol from a requesting node in the network;

means for storing the plurality of search responses;

means for transmitting an aggregate search response to the requesting node;

means for generating an aggregate search result formatted in accordance with the common query protocol including one or more of the plurality of search responses ordered in response to relevance information indicating a ranking of the plurality of search responses; and

means for generating at least one additional search parameter in response to the plurality of search responses being more than a specified number, wherein each of the one or more search responses matches at least one of the at least one additional search parameter.

40. A computer system in a network, comprising:

means for receiving a search response formatted in accordance with a common query protocol from each of a plurality of provider nodes in the network in response to a search query formatted in accordance with the common query protocol from a requesting node in the network;

means for storing the plurality of search responses;

means for transmitting an aggregate search response to the requesting node;

means for generating an aggregate search result formatted in accordance with the common query protocol including one or more of the plurality of search responses ordered in response to relevance information indicating a ranking of the plurality of search responses; and

means for selecting from the plurality of search responses the one or more search responses in response to other relevance information indicating another ranking of the plurality of search responses.

Description

BACKGROUND OF THE INVENTION

1. Field of the Invention

This invention relates to computer networks, and more particularly to a system and method for providing a distributed information discovery platform that enables discovery of information from distributed information providers.

2. Description of the Related Art

It has been estimated that the amount of content contained in distributed information sources on the public web is over 550 billion documents. In comparison, leading Internet search engines may be capable of searching only about 600 million pages out of an estimated 1.2 billion "static pages." Due to the dynamic nature of Internet content, much of the content is unsearchable by conventional search means. In addition, the amount of content unsearchable by conventional means is growing rapidly with the increasing use of application servers and web enabled business systems.

Crawlers currently may take three months or more to crawl and index the web (Google numbers), so that conventional, crawler-based search engines such as Google may best perform when indexing static, slowly changing web pages such as home pages or corporate information pages. Targeted or restricted crawling of headline or other metadata is possible (such as that done by moreover.com) but this limits search ability. Web resources that do not have a "page of contents" or similar index—"deep" web resources—may be more difficult to search, index, or reference by conventional crawler-based search engines. For example, Amazon.com contains millions of product descriptions in its databases but does not have a set of pages listing all these descriptions. As a result, in order to crawl such a resource, it may be necessary—though difficult—to query the database repeatedly with every conceivable query term until all products are extracted. Likewise, many web pages are generated dynamically given information about the consumer or context of the query (time, purchasing behavior, location, etc.), a crawler approach is likely to lead to distortion of such data. In some situations, content may be inaccessible due to access privileges (e.g. a subscription site), or for security reasons (e.g. a secure content site).

Conventional search mechanisms also may be less efficient than desirable in regard to some types of information providers, for example in regards to accessing dynamic content from a news site. A current news provider may provide content created by editors and stored in a database as XML or other presentation neutral form. The news provider's application server may render the content as a web page with associated links using templates. Although the end user may see a well-presented page with the relevant information, for a crawler-type search engine to extract the content of the HTML page it must be programmed to use information about the structure of the page and "scrape" the content and headline from the page. It may then store this content or a processed version for indexing purposes in its own database, and retrieve the link and story when a query matching the story is submitted. This search process is inherently inefficient and prone to errors. In addition it gives the content provider no control over the format of the article or the decision about which article to show in response to a query.

It would be desirable for search mechanism of the web to perform "deep searches" and "wide searches." "Deep search" may find information embedded in large databases such as product databases (e.g. Amazon.com) or news article databases (e.g. CNN). "Wide searches" may reach a large distribution. Moreover, it would be desirable for the search mechanism to efficiently use bandwidth and maximize search speed while avoiding bottlenecks. It would also be desirable for a search mechanism to function over an expanded web covering a wide array of distributed devices (e.g. PCs, handheld devices, PDAs, cell phones, etc.).

SUMMARY OF THE INVENTION

A distributed network search mechanism is described for a consumer coupled to a network to send a search request to and receive a search result from at least one provider coupled to the network in response to its search request. A search request may include a search query. A search result may include a query result. A search request and a search result may be formatted according to a query routing protocol (QRP). A QRP may specify a mark-up language format for communicating search requests, search results, and/or other information between nodes in the network.

A network hub may be configured to implement a search method according to a query routing protocol. The search method may include receiving a search request from a consumer. A network hub may accept search requests only from registered consumers. A network hub may be configured to receive registration requests from consumers. A network hub may be configured to receive registration requests from providers. A registration request may be formatted according to a QRP. A provider's registration request may indicate at least some of the search queries the provider is interested in receiving. The search method may include resolving a consumer's search query from a search request by determining at least one provider that indicated interest in receiving at least similar search queries in its registration request. A network hub may be configured to route a consumer's search query to a provider and may format the search query according to a QRP.

A provider may be configured to receive a search query. A provider may respond with a query result. A provider may be configured to customize its query result. A query result may be formatted according a QRP. The query result may be routed to a network hub. A network hub may be configured to receive a query result from a provider. A network hub may be configured to collate a plurality of query results regarding the same search query. A network hub may be configured to route a query result or collated query results to a consumer as a search result. A search result may be formatted according to a QRP.

A network hub may be configured to route a search request, a search result, or other communication between a consumer and a provider through at least another network hub. A network hub may be configured to resolve a consumer's search query using a query-space. A search request may include an indication of a query-space. A provider registration may include an indication of a query-space. A query-space at least defines a structure for indicating and matching search criteria, and may include a predicate statement. A provider registration may include a query server address to which matching search queries are to be directed.

Resolving a search query may include deriving search criteria from a search query, applying the search criteria from the search query to the search criteria of the query-spaces from provider registrations, and determining which query-spaces from provider registrations suitably match the search criteria from the search query. A search query may be routed to at least a subset of the query server addresses specified by the resolved providers registrations.

A QRP interface may be configured to operate with a consumer or a provider in the network. A QRP interface may be configured as a proxy for a consumer or a provider that do not include a QRP interface to operate with the distributed network search mechanism. A QRP interface may be configured as an interface between a network hub and a consumer or a provider to receive information from that consumer or provider and send it to a network work or to receive information from a network hub and send it to that consumer or provider. A consumer, or a provider may be configured to send information to or receive information from a QRP interface. A network hub may be configured to send or receive information to a QRP interface for a consumer or a provider. A QRP interface may be configured to translate a between consumer or provider specific protocols to a QRP. A QRP interface may be configured to customize a search query or a search result in response to instructions from a consumer or a provider.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 illustrates a network utilizing the distributed information discovery platform according to one embodiment;

FIG. 2 illustrates an architecture for the distributed information discovery platform according to one embodiment;

FIG. 3 illustrates message flow in a distributed information discovery network according to one embodiment;

FIG. 4 illustrates a provider with a query routing protocol interface according to one embodiment;

FIG. 5 illustrates a provider with a query routing protocol interface and a results presentation mechanism according to one embodiment;

FIG. 6 illustrates an exemplary distributed information discovery network including a plurality of hubs according to one embodiment;

FIG. 7 illustrates provider registration in a distributed information discovery network according to one embodiment;

FIG. 8 is a flowchart illustrating message flow in a distributed information discovery network according to one embodiment;

FIG. 9 illustrates an example of several peers in a peer-to-peer network according to one embodiment;

FIG. 10 illustrates a message with envelope, message body, and optional trailer according to one embodiment;

FIG. 11 illustrates an exemplary content identifier according to one embodiment;

FIG. 12 is a block diagram illustrating two peers using a layered sharing policy and protocols to share content according to one embodiment;

FIG. 13 illustrates one embodiment of a policy advertisement;

FIG. 14 illustrates one embodiment of a peer advertisement;

FIG. 15 illustrates one embodiment of a peer group advertisement;

FIG. 16 illustrates one embodiment of a pipe advertisement;

FIG. 17 illustrates one embodiment of a service advertisement;

FIG. 18 illustrates one embodiment of a content advertisement; and

FIG. 19 is a block diagram illustrating one embodiment of a network protocol stack in a peer-to-peer platform.

While the invention is described herein by way of example for several embodiments and illustrative drawings, those skilled in the art will recognize that the invention is not limited to the embodiments or drawings described. It should be understood, that the drawings and detailed description thereto are not intended to limit the invention to the particular form disclosed, but on the contrary, the intention is to cover all modifications, equivalents and alternatives falling within the spirit and scope of the present invention as defined by the appended claims. The headings used herein are for organizational purposes only and are not meant to be used to limit the scope of the description or the claims. As used throughout this application, the word "may" is used in a permissive sense (i.e., meaning having the potential to), rather than the mandatory sense (i.e., meaning must). Similarly, the words "include", "including", and "includes" mean including, but not limited to.

DETAILED DESCRIPTION OF EMBODIMENTS OF THE INVENTION

A system and method for providing a distributed information discovery platform that may enable discovery of information from distributed information providers is described. In an embodiment, in contrast to conventional search engines and exchanges, the distributed information discovery platform does not centralize information; rather it may search for information in a distributed manner. This distributed searching may enable content providers to deliver up-to-the-second responses to search queries from a user or client.

In the distributed information discovery platform, queries are distributed to "peers" in a network who are most likely to be capable of answering the query. The distributed information discovery platform provides a common distributed query mechanism for devices from web servers and small computers.

The distributed information discovery platform may be applied in a wide variety of domains, including, but not limited to: public accessible web search, private networks of trading partners, and interaction between distributed services and applications. In addition to supporting public networks, the distributed information discovery platform may also include support for private networks such as for business-to-business (B2B) networks and extranet applications. Private network support may include quality of service provisioning, security via public key infrastructure and explicit B2B queryspace support. The distributed information discovery platform may also be applied to Peer-to-Peer (P2P) networking, exemplified in programs such as Napster and Gnutella. The distributed information discovery platform may also be applied to other similar networks or combination of networks.

In one embodiment the distributed information discovery platform may include a web front end to a distributed set of servers, each running a P2P node and responding to queries. Each node may be registered (or hard coded in some embodiments) to respond to certain queries or kinds of queries. For example, one of the nodes may include a calculator service which would respond to a numeric expression query with the solution. Other nodes may be configured for file sharing and may be registered to respond to certain queries. A search query on a corporate name may return an up-to-the-minute stock quote and current news stories on the corporation. Instead of presenting only text-based search results, the distributed information discovery platform may return other visual or audio search results. For example, a search query for "roses" may return photo images of roses.

In some embodiments, the distributed information discovery platform may leverage web technologies (e.g. HTTP/XML). In addition to supporting arbitrary XML, the distributed information discovery platform may be integrated with other standard initiatives such as the Resource Description Framework (RDF) for describing metadata and queryspace vocabularies, XML-RPC (XML-Remote Procedure Call (RPC)) for exposing interfaces in a standard manner, Rich Site Summary (RSS) (previously known as RDF Site Summary), Simple Object Access Protocol (SOAP) and Microsoft's NET. These technologies may provide a more familiar environment to developers and webmasters than less common or proprietary protocols. In addition, leveraging such web technologies may simplify a user's task in interfacing to a query routing protocol of the distributed information discovery platform. In one embodiment, a "search button+results" interface item or items may be added to web pages of web sites that may invoke the search capabilities provided by the distributed information discovery platform.

The distributed information discovery platform may provide an abstract query routing service for networks with arbitrary messaging and transport mechanisms. In one embodiment, the distributed information discovery platform may bind with the Web (e.g. XML over HTTP). Note that the distributed information discovery platform may search across heterogeneous communication protocols and systems and present results using any number of different protocols and system. For example, one embodiment of a distributed information discovery system may search JSP-based HTTP systems simultaneously with Perl-based XML systems and Java-based peer-to-peer systems. The distributed information discovery system may then present the results in HTTP-based HTML or according to a peer-to-peer protocol or any other protocol/medium combination.

In one embodiment, the distributed information discovery platform may bind with a peer-to-peer networking environment. In a peer-to-peer networking environment, entities of the distributed information discovery platform (e.g. consumers, providers, hubs, registration services, etc.) may be implemented on peers in the network. Each peer may run instances of the provider, consumer and registration services on top of its peer-to-peer networking core. Each peer may interact with an instance of a hub service, itself running on top of the peer-to-peer networking core. One peer-to-peer networking environment with which the distributed information discovery platform may bind is implemented with a novel open network computing platform for peer-to-peer networks, which may be referred to as a peer-to-peer platform. This peer-to-peer networking environment is described later in this document.

In one embodiment, the distributed information discovery platform may include a Provider Information Service that may include a database and management service for provider information such as contact details, billing information, etc. In one embodiment, the distributed information discovery platform may include a user preferences service that may include a database and management service for end user preferences. Users of the web client may register as users and have the front-end application remember their preferences. In one embodiment, user preferences may be used to provide personalized searching. For example, a user may specify a maximum number of results to be returned.

Embodiments of the distributed information discovery platform may include a monitoring and management tool or tools. Administrators may use the tool(s) to monitor and manage the performance of the distributed information discovery platform. For example, monitoring tools may provide information on the number of searches performed, most popular keywords, most popular clients, most popular providers, etc. Also, performance information on the servers, database uptime etc. may be provided by the tool(s). Management tools may provide the ability to remotely suspend traffic to a provider, for example. For public network applications, "spam" may be addressed in a variety of ways, including comparison of the site registration to an inferred registration, tracking of searches made and results returned and allowing consumer input, such as voting.

Some embodiments of the distributed information discovery platform may be used for two complementary search types: wide and deep. The concept of the expanded web covers both wide search of distributed devices (e.g. PCs, handheld devices, PDAs, cell phones, etc.) and deep search of rich content sources such as web servers.

In one embodiment, the distributed information discovery platform may be used to provide "wide search" on the web. Within the context of wide search, the distributed information discovery platform may provide an efficient mechanism for distributing queries across a wide network of peers. The distributed information discovery platform may use a series of "hub" peers each of which handles the queries for a group of peers. Each hub peer may specialize in an attribute such as geography, peer content similarity or application. Hub peers may forward queries to other hub peers either if they cannot satisfy the query or if it is desirable to expand the search to the widest number of peers possible.

In one embodiment, the distributed information discovery platform may be used to provide "deep search" on the web. "Deep search" may find information embedded in large databases such as product databases (e.g. Amazon.com) or news article databases (e.g. CNN). In one embodiment, rather than crawling such databases, indexing and storing the data, the distributed information discovery platform may be used to determine which queries should be sent to such databases and direct these queries to the appropriate database provider(s). The database provider's own search capabilities may be employed to respond to the query through the distributed information discovery platform. Thus, the resulting search results may be more up-to-date and have wider coverage than a set of conventional crawler search engine results.

The ability to search recently updated information may make the distributed information discovery platform better suited for "deep search" than existing crawler-based search engines. The distributed information discovery platform may leverage remote access or public search capabilities provided by information providers. Furthermore, under the distributed information discovery platform, a provider that wishes to restrict remote access may still allow searching and control how content is searched by registering with a distributed information discovery network. The distributed information discovery platform may specify a common query routing protocol which may give both parties more flexibility and control of the exchange of data, which may improve search efficiency in some embodiments.

Application Ser. No. 60/308,932 entitled "TRUST MECHANISM FOR A PEER-TO-PEER NETWORK COMPUTING PLATFORM" by William J. Yeager and Rita Y. Chen is hereby incorporated by reference.

FIG. 1 illustrates a network that utilizes the distributed information discovery platform according to one embodiment. The distributed information discovery platform may be applied to create a distributed information discovery network having three main types of participants: providers 120, consumers 140, and hubs 100. In many applications, a program or node may act as both provider 120 and consumer 140. A network may encompass a cloud of machines. Physically, a provider 120 or a consumer 140 may be, for example, an individual computer, set of computers, computing process, or a web service. In one embodiment, providers 120 and consumer 140 may be any peer within a network, including peer-to-peer platform peers running a distributed information discovery platform or HTTP peers adapted to a query routing protocol. A hub may be implemented on one or more machines or processes, and a program acting as a provider or a consumer may also function as a hub. The term "computer" is not limited to any specific type of machine and may include mainframes, servers, desktop computers, laptop computers, hand-held devices, PDAs, telephone or mobile phones, pagers, set-top boxes, or any other type of processing or computing device.

Consumers 140 may query the distributed information discovery network and receive responses from providers 120. A consumer 140 may be defined as anything that makes requests in the network. A consumer 140 may be, for example, a peer in a peer-to-peer network or a web site with an HTTP client interface to the network. In one embodiment, the query may be sent to a hub 100 nearest to the consumer 140, which routes the query to all interested providers 120. "Nearest" in this sense does not necessarily imply geographical nearness, but instead refers to a hub 100 that is at the fewest "jumps" (shortest route) to the consumer 140 on the network. In one embodiment, the distributed information discovery platform may include information, for example its location in the network, regarding a hub with which a consumer should communicate in the distributed information discovery network.

A network routing system, referred to as a hub 100, may handle query and response routing in the network. A hub 100 may act as an access point that may provide virtual access to a portion of or the entire distributed information discovery network. Providers 120 and consumers 140 may contact the network through a specific hub 100 implemented on one or more machines. In some embodiments, providers 120 and consumers 140 may contact different hubs 100. Hubs 100 may facilitate efficient query routing over the network by handling message routing between consumers 140 and providers 120. In one embodiment, a hub 100 may include a router 104 that handles the routing of queries to providers 120. In one embodiment a hub 100 may include a router 104 that handles the routing of responses to consumers 140. The hub 100 may determine one or more providers 120 of which the hub 100 is aware (e.g. that have registered with the hub 100) and that may be qualified to process a received query. In one embodiment, a hub 100 may include a resolver 102 which may handle the determination of qualified providers 120.

In some embodiments, queries may be resolved by a resolver 102 in the network by matching query terms to registration terms. In some embodiments, the resolver 102 may use simple keyword based matching of query terms to registrations. In other embodiments, the resolver may be extended, for example, to allow for category based matching of terms to registrations and/or adaptive learning of provider performance, e.g. learning which providers return relevant results given certain kinds of queries. Providers 120 whose registration terms match the query terms may be returned by the resolver 102. The hub 100 may include metadata associated with the providers 120, including the provider descriptions registered with the hub 100. This metadata may be used to determine the qualified provider(s) 120. The hub 100 then may send the query to the provider(s) 120 it has determined to be qualified. Each provider 120 that receives the query may process the received query and send one or more responses to the hub 100. The hub 100 may receive the responses and route them to the consumer 140 that initiated the query.

A provider 120 may be defined as anything that responds to requests (queries) in the network. A provider 120 may be, for example, a peer in a peer-to-peer network or a Web server such as cnn.com. The distributed information discovery platform allows information providers 120 to publish a description of queries that they are willing to answer. In one embodiment, each provider 120 may register a description of itself on the distributed information discovery network. In one embodiment each provider 120 then waits for requests matching information in the description. In one embodiment, providers 120 may register by sending registration information to the hub 100. The registration information may include metadata describing the types of queries that a provider 120 may be able to respond to. In one embodiment, the registration information may be maintained in a registration repository that may include registration information for a plurality of providers 120. In one embodiment the hub 100 has access to the registration repository.

In one embodiment provider registrations may be meta-data indexes. Registration information for a provider 120 may include queryspaces that define which queries the provider 120 may respond to. The registration, in one embodiment, may include an XML-based encoding of a logical statement characterized by a queryspace, optionally characterized by a schema. In one embodiment, if no schema is specified a default schema for general keyword matching may be used. For example, a user may send a search query to a distributed information discovery routing system. The query may be compared to the registrations (e.g. meta-data indexes). In one embodiment, the registrations may be stored in XML format describing a conjunctive-normal logic. Queries are then routed to providers matching the query.

In some embodiments, users and end applications (consumers 140) may present queries to a distributed information discovery network as arbitrary XML. Schema selection may be performed by HTTP header specification, in some embodiments. In one embodiment, queries presented by consumers 140 may adhere to specific queryspaces. In some embodiments, queries may be routed to the appropriate provider 120 by sending requests (e.g. XML requests) over HTTP. A router 104 may send the requests and await responses. In some embodiments, the router 104 may continually monitor providers to determine availability and reliability. Providers 120 may respond to queries in, e.g., arbitrary XML that may include links to any results they have in their site.

In some embodiments matches, results and their ordering may be determined according to relevance. The relevance may be specified by the user or alternatively may be a pre-defined relevance. In some embodiments the distributed information discovery network may perform some tailoring of the responses to search queries, for example by enabling providers to select the information to send in response to search queries or by ranking the results based on information from any of the providers. In one embodiment, the distributed information discovery network may not perform any presentation of the responses from providers 120. In this embodiment, the consumer may include a front end to perform such presentation, e.g. either as a web page or as a client side user interface. In one embodiment, the distributed information discovery network may collate results from providers 120, perform ranking on the results with respect to the query and present them in HTML, for example. Thus, a general application or user (consumer 120) may be able to query a distributed information discovery network and act on the responses as it sees fit. For example a music file sharing application may receive results and sort them according to file size/connection rate. In some embodiments, the links are provided to the information matching the queries.

In addition to functioning as a "meta-search" engine, the distributed information discovery platform may include support for an open protocol for distributed information routing. This protocol for distributed information routing may be referred to as a query routing protocol (or QRP). The query routing protocol may be used in defining queries, responses and registrations. The query routing protocol may allow both structured, lightweight and efficient query message exchange. In one embodiment, the query routing protocol may be implemented in XML. The query routing protocol may define mechanisms for sending and responding to queries in the network, in addition to mechanisms for defining metadata for nodes in the network. In one embodiment, this query routing protocol allows information providers to publish a description of queries that they are willing to answer. Information consumers may submit queries to the network, which routes each query to all interested providers. The query routing protocol may allow participants in the network to exchange information in a seamless manner without having to understand the structure of the presentation layers. Embodiments of the query routing protocol may be based on existing open standards, including markup languages such as XML (eXtensible Mark-up Language) and XML Schema. In addition, the query routing protocol may be encapsulated within existing protocols, such as HTTP (HyperText Transfer Protocol).

In some embodiments, the query routing protocol of the distributed information discovery platform may provide an interface designed for simplicity. For example, a minimally-conforming client implementation may be built in one embodiment using existing libraries for manipulating XML and sending HTTP messages. A minimally-conforming server implementation may be built in one embodiment with the above tools plus a generic HTTP server.

The query routing protocol of the distributed information discovery platform may provide structure. For example, in one embodiment, queries on a distributed information discovery network may be made using XML messages conforming to a particular schema or queryspace. Since providers may have widely differing kinds of content or resources in their datastores, the query routing protocol may be used to define queryspaces that may be used to define the structure of queries and the associated registration information for a provider 120. In one embodiment, queryspaces may define the structure of a valid query that a provider 120 can process. In one embodiment, queryspaces may be implemented in XML. In such an embodiment, information providers may register templates describing the structure of queries to which they are willing to respond.

The query routing protocol of the distributed information discovery platform may provide extensibility. In some embodiments, arbitrary schemas or queryspaces may be used on a distributed information discovery network. In such embodiments, there may be no need for centralized schema or queryspace management. Thus, ad hoc collaboration may be simplified.

The query routing protocol of the distributed information discovery platform may provide scalability. For example, in one embodiment, a distributed information discovery network may support millions of publishers and consumers performing billions of transactions per day. In some embodiments, sophisticated implementations may take advantage of advanced connection-management features provided by lower-level protocols (e.g. HTTP/1.1).

The following describes one embodiment of a query routing protocol that may be used in embodiments of the distributed information discovery platform. In this embodiment, the query routing protocol may include several components. One component may be a query request. Another component may be a query response. A component may be a registration.

Registrations may be structured to delineate the different information included by a provider in that message. For example, a registration body may be enveloped within the <register> and </register> tags. A query server, i.e. the URL or Pipe ID of the provider to send the queries to is specified within <query-server> and </query-server>. A "predicate", i.e. the logical statement which the queries must match to be routed to this provider is enveloped within tags <predicate> and </predicate>. A predicate may include the queries that will be matched by this provider, each enveloped within <query> and </query> tags. Each predicate may contain multiple <query> envelopes. A query body may contain arbitrary XML as long as it matches the namespace that matches the specified query-space for this provider. For example, query bodies containing the terms "html", "java" or "xml" would be routed to provider "http://abcd.com/" which may have registered those terms when registering as a provider as follows:


	<?xml version='1.0'?>
	<register xmlns="http://abcd.com"
	query-server=http://abcd.com/search>
	<predicate>
	<query><text>html java xml</text></query>
	</predicate>
	</register>

An example registration for abcd.com may look like this:


	<?xml version='1.0'?>
	<register xmlns="http://abcd.com"
	xmlns:b="http://bigbookseller.com/search"
	query-space="http://bigbookseller.com/search">
	<query-server>http://abcd.com/search</query-server>
	<predicate>
	<query>
	<b:author>John Doe Jane Doe</b:author>
	<b:title>Foos Gadgets Widgets</b:title>
	</query>
	</predicate>
	</register>

Query messages may be structured to indicate which portions are queries and which include other information. For example, a default namespace may be specified by a URI such as "http://abcd.org/search". A query message may be contained within the envelope <request> . . . </request>. A query unique ID may be specified in a uuid attribute of the <request> tag. A query space may be specified within the tags <query-space> and </query-space>. An actual query data may be enveloped within the tags <query> and </query>. Query data may be arbitrary XML within a namespace that matches "http://abcd.org/search/text", which includes the tag <text> to specify free text, or within any other namespace specified by the <query-space> definition. Generally any envelop or structure may be used provided it adequately identifies information needed to define query information between members of the search network. In one embodiment, each query request message includes the <request uuid="uuid details">, <query-space>, and <query> tags. In one embodiment, although <query-space> defines a name for the type of query that is being performed, a name space may be used with the same value as that specified in <query-space> for all the queryspace-specific tags. One embodiment may use a full XML schema framework for defining and validating queryspaces.

Response messages may be structured to indicate which portions are responses and which include other information. For example, a default name space may be "http://abcd.com/search". A response message may be enveloped within the <response> and </response> tags. A body of the response may be arbitrary XML as long as it corresponds to the specified queryspace and corresponding namespace, e.g. the queryspace "http://abcd.com/search" includes the <text> tag. Generally any envelop or structure may be used provided it adequately identifies information needed to define response information between members of the search network.

The following is an example of the format of a query for the term "foo":


	<?xml version='1.0'?>
	<request xmlns="http://abcd.com/search"
	xmlns:t="http://abcd.com/search/text"
	uuid="1C8DAC3036A811D584AEC2C23">
	<query><t:text>foo</t:text></query>
	</request>

The following is an example of a response to this query:


	<?xml version='1.0'?>
	<response xmlns="http://abcd.com/search">
	<text>Hi, I'm a peer-to-peer platform peer</text>
	</response>

A more complex example may be:


	<?xml version='1.0'?>
	<request id="1C8DAC3036A811D584AEC2C23"
	query-space="http://bigbookseller.com/js"
	xmlns="http://abcd.com/search"
	xmlns:books="http://bigbookseller.com/search">
	<query>
	<b:author>John Doe</b:author>
	<b:title>Widgets</b:title>
	</query>
	</request>

In this example the query space is defined as "http://bigbookseller.com/search" and the namespace "books" matches the URI for this query space. The query specifies that the "author" within the name space "books" should be "John" or "Doe" or that the title should contain "Widgets".

An example of a response by abcd.com may be:


	<?xml version='1.0'?>
	<response xmlns="http://abcd.com/search"
	xmlns:b="http://bigbookseller.com/search"
	query-space="http://bigbookseller.com/search">
	<b:authors>John Doe, Jane Doe</b:authors>
	<b:URL>
	http://www.abcd.com/obidos/ASIN/0201310082
	</b:URL>
	<b:title>Foos, Gadgets and Widgets</b:title>
	<b:price>$39.95</b:price>
	<b:abstract>
	A definitive technical reference for foos, gadgets and
	widgets, written by the inventors of the technologies.
	</b:abstract>
	</response>

In addition, request messages may contain optional attributes. These may be contained inside request tags. If unspecified, defaults attributes may be assumed. Optional attributes may include: "max-hits-per-provider" indicating a number of hits expected from a provider; "flushafter" to indicate to flush the output stream to the client after receiving responses from a certain number of providers; "queryuuid" to indicate a unique id of the query; "querylifetime" to indicate a length of time during which the query is valid; or "maxfanout" to indicate a maximum number of providers to which to forward the query. For example, a tag may be: <request flushafter=5 providerhits=2 timeout=2>.

An architecture for the distributed information discovery platform is shown in FIG. 2, according to one embodiment. In one embodiment, a consumer 140 may provide users an access point to a distributed information discovery network. A consumer such as consumer 140A may include a consumer query request protocol interface 142. A QRP interface may be a stand alone application, a component of a distributed information discovery platform, a script capable of parsing requests and generating an appropriately formatted response, or any hardware or software configured to include at least functionality for translating to or from query response protocol data. The consumer QRP interface 142 may send queries written in the query request protocol to the hub 100 for query resolution and routing. After sending a query, the consumer QRP interface 142 may await responses from providers. In one embodiment, the queries may be received by a hub consumer QRP interface 108 of router 104. In one embodiment, the consumer QRP interface 142 may also perform formatting of the responses for presentation to the end user or application, which may include ordering or otherwise organizing the responses. In one embodiment, the formatting or ordering of the responses may be in response to instructions received from the consumer or provider. In one embodiment, consumers 140 may also include a front end or user interface (e.g. a web user interface) to the hub (e.g. the router and/or resolver). In one embodiment, a consumer 140 may include a mechanism for ranking and presentation of query results. In one embodiment, this mechanism may be a component of the consumer QRP interface 142. Ranking methodology may be implicit in each queryspace, and may be returned as part of each response in some embodiments. Some ranking schemes may require third-party involvement.

In one embodiment, consumers such as consumer 140C may not include a consumer QRP interface 142. These consumers may use a consumer proxy 110 to interface to the functionality of the hub 100. The consumer proxy 110 may perform translation of queries formatted in one or more query protocols supported by the consumers 140 into queries in the query routing protocol. These queries may then be sent to the hub 100 for resolution and routing. In one embodiment, the queries may be received by a hub consumer QRP interface 108 of router 104. The consumer proxy 110 may also perform translation of query responses formatted in the query routing protocol into one or more protocols supported by the consumers 140. As shown, one or more consumers 140 may interface with the consumer proxy 110.

In one embodiment, a provider such as provider 120A may include a provider query request protocol (QRP) interface 122 that may accept queries from the hub 100 in the query routing protocol and respond to the queries with query responses in the query routing protocol. The provider QRP interface 122 may perform translation of queries into provider-specific requests. In one embodiment the QRP interface 122 may include an indexing and/or searching interface and may be configured to perform indexing and searching itself. In one embodiment, the provider QRP interface 122 may not perform any indexing or searching itself, but rather may call the appropriate indexing and/or searching interface of the provider 120, for example, a database search engine. In this embodiment, the provider QRP interface 122 may, if necessary, translate the queries from the query request protocol into a protocol that may be used by the appropriate indexing and/or searching interface of the provider 120. The provider QRP interface 122 may also, if necessary, translate the query responses from the protocol used by the appropriate indexing and/or searching interface of the provider 120 into the query request protocol. A provider QRP interface 122 may be, for example, a small modification of an existing search engine script (Java Server Page (JSP), Perl etc.) so that queries from a distributed information discovery network can be applied to the provider's search engine.

Provider proxy 114 may perform translation of queries formatted according to the query routing protocol to specific search engine formats for a provider 120 such as provider 120C. Provider proxy 114 may also perform translation of responses formatted according to the specific search engine formats into responses formatted according to the query routing protocol. A provider proxy 114 may be used, for example, if a provider 120 does not run its own provider QRP interface 122B but does allow access to its own search engine.

Hub 100 performs the routing of queries from consumers 140 to providers 120. The hub 100 accepts queries, resolves those queries to the appropriate providers 120 and then manages the routing of the queries to the providers 120. The hub 100 then may collate the results received from one or more providers 120 and send the results back to the requesting consumer 140 in a query response.

In one embodiment, rather than sending the results back to the consumer 140 in a query response, the results may be provided to the client by other means. For example, the query message may include an email address or addresses to receive the results. After receiving and collating the results, the hub 100 may email the results to the email address(es) specified in the query message. In one embodiment, the hub 100 may also receive queries in email messages from consumers. As another example, the hub 100 may post the results to a URL specified in the query message. Alternatively, the provider 120 may provide the results directly to the consumer 120 rather than routing the results through the hub 100. The query message may include information that allows the provider 120 to provide the results directly to the consumer 120. For example, the query message may include an email address and/or a URL for the consumer 140, and the provider 120 may email the results to the specified email address or send the results directly to the URL specified in the query message.

A hub 100 may comprise a router 104 that may provide a portion of the functionality of the hub 100. The router 104 may route queries to providers 120, manage query connections, collate results and return responses to consumers 140. A hub may also comprise a resolver 102 that matches queries to providers 120. Provider information 106 may include one or more registration files comprising metadata specified by the providers 120 during registration.

In one embodiment, the resolver 102 may be based on a full text search engine. For example, the core components may be adapted from the Lucene search engine (http://www.lucene.com) written using Java. In one embodiment, the resolver 102 may index all tags and text in the registration files. A reverse index may be created which maps query terms to providers. For efficiency, the resolver may create separate indices for each queryspace.

In one embodiment, a provider 120 may accept queries in the query routing protocol directly from consumers 140 without the queries being routed by the hub 100. In one embodiment, a provider 120 may also return responses to queries directly to consumers 140 without routing the responses through the hub 100.

In one embodiment, a distributed information discovery system may be implemented as a series of distinct web services. Each of the router, resolver, proxies, and QRP interfaces may run independently. In one embodiment, these web services may be implemented as Java Servlet classes referencing additional Java classes for core functions. For example, in a web embodiment, each of the router, resolver, proxies, and QRP interfaces may be implemented on a web-accessible server or servers. Also, a distributed information discovery network may include multiple different routers, resolvers, proxies, and QRP interfaces. One router or resolver may register with another router or resolver. Or one distributed information discovery system implementing a router, resolver, proxies, and/or QRP interfaces may register with another such system. For example, a distributed information discovery routing system may register providers of information concerning outdoor recreation. Another distributed information discovery routing system having provider registrations for boating may register with the first system. In some embodiments, a distributed information discovery system may be implemented on different peers in a peer-to-peer network, or other networks.

In one embodiment, a database used by any of the above components may be a database that provides persistency, such as a GOODS (Generic Object Oriented Database System) database. GOODS is an object-oriented fully distributed database management system (DBMS) using an active client model. Other databases based on other DBMSs may also be used.

Using the proxies and QRP interfaces described above, the distributed information discovery platform may offer a unique technology by enabling search across heterogeneous communication protocols and systems and presenting those results using any other protocol and system. An example of this is the distributed information discovery platform's ability to search Java Server Page (JSP) based HTTP systems simultaneously with Perl-based XML systems and Java-based peer-to-peer protocol systems. The distributed information discovery platform may also provide a mechanism for presenting those results in HTTP-based HTML, a peer-to-peer protocol, or any other protocol/medium combination.

The consumer and provider proxies and QRP interfaces may serve as adaptors for multiple data sources to plug into a standardized interface for distributed deep search. In one embodiment, the distributed information discovery platform is an XML-based request/response system. By using an XML-based messaging format, the distributed information discovery platform may enable powerful and easily implemented deep web searches. Participants in the distributed information discovery platform network need only apply fairly common and available facilities to adapt their system as a network provider. The XML nature of the response messages additionally expands the scope of a provider's ability. Applications other than web browsers may manipulate the responses for different purposes such as determining an average price or a current demand based upon availability.

To be a network provider, participants may include a provider QRP interface 122 that may be tailored for a provider's specific system. A provider QRP interface 122 may parse or translate a query routing protocol request from the distributed information discovery network, query a provider back end 180 to get appropriate data 182, and then generate a response and send it back to the distributed information discovery network according to the query routing protocol. In one embodiment a QRP interface may determine whether a query is recognizably formatted, contains an illegal query or result, would access restricted information, or otherwise cannot validly be processed, and may return an error message or code or similar indication. In one embodiment, the distributed information discovery platform may provide one or more generic QRP interfaces that may be used as examples that illustrate how to use a specific language to accept requests and/or generate responses. The distributed information discovery platform may also provide one or more QRP interfaces that plug into existing, freely available systems.

In one embodiment, queryspaces may be defined within the distributed information discovery platform that enable providers 120 to do more than return links to web pages. For example, rather than querying a database, a QRP interface 122 may compute a price for a particular service based on demand in real time. The QRP interface 122 may generate data, or may cause another application to generate data on demand. An example may be an auction system for spare CPU cycles, where a client would query various providers 120 for CPU time. The providers 120 may generate a price based on current availability. In one embodiment, the distributed information discovery platform may include a result presentation mechanism that may perform computations on and/or presentation formatting of results data.

There may be some differences in some of the internal mechanisms of embodiments that bind to different networks. In general, the query routing protocol and the resolution mechanism may be the same or similar in the different embodiments. The routing mechanism and the client interfaces in the different embodiments, however, may be implemented at least partially differently to support the different network types.

FIG. 3 illustrates message flow in a distributed information discovery network according to one embodiment. An application on consumer 140 may find information providers 120 to respond to a particular query by sending the query into the network via a specific access point (hub 100). In one embodiment, consumer 140 may send the query to router 104 of hub 100. Router 104 may then send the query to resolver 102. In one embodiment, queries may conform to the query routing protocol. In one embodiment, queries are markup language (e.g. XML) messages with essentially arbitrary structure. In this embodiment, there are no restrictions on what tags may be used in queries.

Resolver 102 may determine one or more providers 120 which may receive the query. One or more information providers 120 may have previously registered with hub 100 by sending registration messages each including one or more queryspaces for the particular provider 120. In one embodiment, information from the registration messages, including queryspace information, may be maintained in provider information 106. Provider information 106 may be a file or database of files including registration information for one or more providers 120. Resolver 102 may index and search provider information 106 for queryspaces that match the query. For a provider 120 to be selected to receive the query, the queryspace specified in the query must match a queryspace of the provider 120. Also, the path predicate specified in the registration message must select a non-empty set of nodes in the query.

After determining the one or more providers 120 to receive the query, resolver 102 may provide a list of the selected one or more providers 120 to router 104. Router 104 may then send the query to each of the selected one or more providers 120. Once an information provider 120 receives the query, it composes a response and sends it back to the router 104 of hub 100. Hub 100 may receive one or more responses from each provider 120 that was sent the query. Router 104 may then forward the received responses to the consumer 140, and thus to the querying application. In one embodiment, the hub 100 does not evaluate competing relevance rankings. In one embodiment that task is left to the querying application.

In one embodiment, the hub 100 may collate the responses received from one or more providers 120 prior to sending them to the consumer 140. In this embodiment, consumers 120 are not required to listen for asynchronous responses. Collation may also provide security benefits. For example, collating responses may help prevent distributed denial-of-service attacks based on spoofed queries. Also, the distributed information discovery network may be used to establish peer-to-peer connections.

In one embodiment, the consumer 140 may connect to the resolver 102 initially to request a set of providers 120 to be targets of a query, and then sends this list of providers 120 to the router 104, which manages the query routing from the consumer 140 to the providers 120, and which also returns the results to the consumer 140 (i.e. Consumer->Resolver->Consumer->Router->Providers->Router->Consumer).

FIG. 4 illustrates a provider 120 with provider QRP interface 122 interfacing to a provider search engine backend 180 according to one embodiment. In one embodiment, a provider QRP interface 122 may serve as an adaptor to the query routing protocol. In one embodiment using a common protocol based on a technology such as XML, fairly common and available facilities may be used to create a provider QRP interface 122 to serve as an adaptor between a provider backend and the distributed information discovery network.

Thus, to be a network provider, participants may include a provider QRP interface 122. A provider QRP interface 122 may be tailored for a provider's specific system. A provider QRP interface 122 may parse or translate a query routing protocol request from the distributed information discovery network, query a provider back end 180 to get appropriate data 182, and then generate a response and send it back to the distributed information discovery network according to the query routing protocol. A provider QRP interface 122 may be a stand-alone application or alternatively a script capable of parsing the requests, gathering data and generating an appropriately formatted response.

In one embodiment, providers or back-end systems may send response messages to the provider QRP interface 122 using the Rich Site Summary (RSS) protocol as a default protocol. RSS is an XML protocol designed for site summaries. Using RSS may provide a common formatting standard of the responses, removing the need to handle custom HTML or other custom protocols being returned from providers. In one embodiment, provider proxies are configured to use RSS.

In one embodiment, QRP interfaces may support queries wrapped as XML-RPC (XML Remote Procedure Call (RPC)) requests. XML-RPC is a protocol (which forms the basis of Simple Object Access Protocol (SOAP)) for invoking server side methods over XML. In other embodiments, QRP interfaces also support HTML or other formats for data transmission or data gathering.

FIG. 5 illustrates a provider 120 with provider QRP interface 122 interfacing to a provider search engine backend 180 according to one embodiment. In this embodiment, a result presentation mechanism 190 is shown that may enable providers 120 to do more than return links to web pages. For example, result presentation mechanism 190 may take the search results in the response message from search engine 180 and tailor the results into a presentation format such as a markup language document. This markup language document may be sent to the provider QRP interface 122, which may package the document in a QRP response and send it to the consumer 140. In one embodiment, QRP interface 122 includes result presentation mechanism 190. As another example, a result presentation mechanism 190 may compute a price for a particular service based on demand in real time. As an example, in an auction system for spare CPU cycles, a consumer 140 may query various providers 120 for CPU time. The providers 120 may generate a price based on current availability.

The distributed information discovery platform may be used for augmenting standard search engines that index statically available web pages. Standard web pages are useful mainly for web browsers. Other devices, such as wireless communications devices, may benefit from searches that expose relevant data. The distributed information discovery platform may provide the ability to collect queries and provide results with meaningful relevance to a wide variety of information consumers and producers. The distributed information discovery system may use a dynamic data collection methodology that is independent of an information provider's presence on the World Wide Web, for example. An information provider may use a provider QRP interface 122 to function as an adapter and handle incoming queries, provide a registration that defines which services and information are available for what devices (e.g. cell phones, PDAs, etc.), and use a result presentation mechanism 190 to tailor results for presentation on the particular devices. For example, a cell phone may be used to find open service stations and compare prices, or restaurants to compare menus. A consumer QRP interface may be integrated in the cell phone, or may be accessible from the cell phone device to handle queries and responses, tailoring results for presentation on the cell phone. The consumer QRP interface may also similarly be integrated in other mobile or portable devices, and computers generally.

For providers 140 that do not run an adapter for the distributed information discovery platform, a hub 100 may run a provider proxy 114 as illustrated in FIG. 2. A provider proxy 114 may perform translation of queries formatted according to the query routing protocol to specific search engine formats for a provider 120. Provider proxy 114 may also perform translation of responses formatted according to the specific search engine formats into responses formatted according to the query routing protocol. In one embodiment, the provider proxy 114 may perform off-line spidering and indexing of the providers 140 and respond to queries as a standard search engine would (this could be considered an open search indexing service). In another embodiment, the provider proxy 114 may perform translation of queries formatted according to the query routing protocol to specific search engine formats for a provider 120, and may also perform translation of responses formatted according to the specific search engine formats into responses formatted according to the query routing protocol.

FIG. 6 illustrates an exemplary distributed information discovery network including a plurality of hubs 100 according to one embodiment. Each hub 100 may support one or more providers 120 and/or consumers 140 which may use the hub 100 as an access point to the distributed information discovery network. As shown in node 180, a node on the network may include instances of both a consumer 140 and a provider 120. In one embodiment, the distributed information discovery platform may support nodes comprising one or more consumers 140, one or more providers 140, and/or one or more hubs 100.

In one embodiment, the distributed information discovery network may include one or more hubs 100 that each may support a particular type of application or specialist domain. For example, a web site might run a hub 100 as a vertical aggregator of content pertaining to Java programming. Its providers 120 may include other sites with content focused on Java. However, the web site may also send queries out to a different hub 100 running on a more general technology news site whose providers 120 may include sites such as CNet or Slashdot, for example. As another example, in a peer-to-peer network, hubs 100 may be used to group together peers with similar content, geography or queryspaces. Each peer within the network may interact with the hubs 100 using its appropriate service(s) (e.g. provider, consumer, and/or registration services).

FIG. 7 illustrates provider registration in a distributed information discovery network according to one embodiment. Information providers 120 may register themselves within a distributed information discovery network. To register, a provider 120 may contact a hub 100 with a registration message. The registration message may conform to the query routing protocol (QRP). In one embodiment, a provider 120A may include a provider QRP registration interface 124 that is operable to send a registration message to the hub 100. In one embodiment, hub 100 may include a QRP registration interface 112 that may be configured to receive registration messages from providers 120. Provider QRP registration interface 124 may also maintain a registration file for the provider 120A. In one embodiment, the distributed information discovery platform may include a registration service 160 that may provide a QRP registration interface to hub 100 for providers 120 that do not include a provider QRP registration interface 124.

Providers 120 may specify the type of queries they wish to receive in a registration file that may be provided to a hub 100 at provider registration. In one embodiment, a registration file may be an XML document comprising metadata about the information that the provider 120 wishes to expose. This file may encode the type and structure of queries, queryspaces and response formats compatible with provider 120. A QRP interface may use the type and structure information in the file to encode queries, queryspaces and responses in formats compatible with provider 120.

The registration file can be thought of as an advertisement of the provider's metadata and its structure. The registration file may include information specifying one or more of several items. For example, a provider's query server endpoint may be included. If this is a peer-to-peer network implemented using the peer-to-peer platform described herein, the endpoint may be a pipe identifier or advertisement. In the web domain, this may be a CGI script which is capable of processing the query request protocol request messages and responding with a query request protocol response. In other embodiments, the endpoint may be a URL. Queries which match one of the provider's predicates may be posted to this endpoint. The file may include a queryspace of the queries this provider will accept. In one embodiment, this may be specified as a queryspace URI (e.g. URL). When queries are posted to this queryspace, the query may be checked against the provider's predicates for matches. The file may include a response format that the provider is capable of responding in. The response format may be specified as a URI to an XML schema. The file also may include a structure and content of the queries the provider is interested in receiving, specified in predicate form. In one embodiment, a set of predicates may define the structure and content

In one embodiment, a registration message may include the following tags:


	<register>...</register>	- tags identifying this as a registration
		document
	<predicate>...</predicate>	- tags enveloping a predicate

The following is an example registration document according to one embodiment:


	<register>
	<queryspace>http://www.abcd.com/opensearch</queryspace>
	<query-server>http://www.efgh.com/search.jsp</
	query-server>
	<predicate>baba ghannouj ghannoush ganoush</predicate>
	</register>

This example registers a provider 120 with a queryspace. It also registers one predicate that will direct any query containing any of the words "baba", "ghannouj", "ghannoush" or "ganoush" to the provider's query server running at http://www.efgh.com/search.jsp. This matches any query containing the particular keywords.

As another example, consider the following registration:


	<?xml version='1.0'?>
	<register xmlns="http://abcd.org/search"
	xmlns:b="http://bigbookseller.com/search"
	query-space="http://bigbookseller.com/search"
	query-server=http://littlebookseller.com/exec/search>
	<predicate>
	<query>
	<b:author>
	John Doe Jane Doe
	</b:author>
	<b:title>
	Foobar Gadgets Widgets
	</b:title>
	</query>
	</predicate>
	</register>

This registers a provider 120 with the text queryspace, specified by http://bigbookseller.com/search. This registration registers the provider for the following queries: any query containing "John Doe" or "Jane Doe" in the <author> field and any query containing "Foobar", "Gadgets", or "Widgets" in the <title> field.

Queries matching these conditions may be directed to the query server running at http://littlebookseller.com/exec/search. Predicates may be much larger than this exemplary predicate, and may also contain more complex structure.

In some embodiments, if the provider 120 does not specify a queryspace, a default queryspace may be registered for the provider 120. In such an embodiment, queries failing to indicate a queryspace may be assumed to be of the default queryspace.

In one embodiment, a provider may be registered using a user interface in which keywords may be typed or pasted. In one embodiment, the user interface may be a Web page. In one embodiment, providers may be able to choose from a list of categories in addition to choosing keywords for their registrations. These categories may reflect the contents of open directories such as dmoz.org and some common news sources (e.g. CNN). For example, the top level of dmoz may be used as a pull down list or menu of categories from which providers may choose. In one embodiment, further specialization in categories may be provided—e.g. for News, providers may choose News->Tech News. In one embodiment, a recursive menu system may be used—e.g. a provider picks News, then presses submit, then picks Tech News and so on. The category data may be updated as needed—e.g. daily for news, weekly for other categories.

In one embodiment, providers may edit their registration information via a user interface (e.g. web page) or a web form, or alternatively submit a replacement/addition to their registration. In one embodiment a QRP adapter may monitor or log queries, results, number of hits, searches, results, etc. or generally the information passing through the QRP adapter. In one embodiment, a user interface may be provided through which providers may view the results of searches and hits performed by consumers—e.g. how many searches resulted in their entry being returned, how many users clicked through, etc. In one embodiment, a user interface may be provided through which providers may monitor and/or control the number of queries sent to them and also to throttle traffic (e.g. turn it off) if necessary. In some embodiments, a QRP interface may be able to access a registration file, for example to read at least part of the registration document or to write to replace or to add to at least part of the registration document.

An embodiment may include a site analysis tool that may be used for building registrations for sites that do not know how to or that do not desire to build their own registration. The site analysis tool may be available as an option during registration (for example, "build me a registration file" with a turn around of 24 hours or so), and may allow the provider to enter one or more initial keyword starting points. The site analysis may produce a queryspace from the information available through a site to reflect the kind of query to which the site may respond. In one embodiment the site analysis tool is part of a QRP interface. In one embodiment the QRP interface is a proxy to a provider. The tool site analysis tool may query, crawl, spider, index, or otherwise access or interact with the site to determine the type of information available from the site.

FIG. 8 is a flowchart illustrating message flow in a distributed information discovery network according to one embodiment. An application on a consumer may find information providers to respond to a particular query by sending the query into the network via a specific hub. In one embodiment, queries may conform to a query routing protocol. In one embodiment, a consumer QRP interface is configured to produce queries that conform to a query routing protocol. In one embodiment, queries are markup language (e.g. XML) messages with essentially arbitrary structure. In this embodiment, there are no restrictions on what tags may be used in queries.

The consumer may send the query to the hub as indicated at 300. In one embodiment, a router on the hub may receive the query. In one embodiment, a query routing protocol interface of the consumer may translate the query from a protocol understood by the consumer to the query routing protocol before sending the query to the hub. As indicated at 302, the hub may resolve the query to determine one or more providers that may want to process the query. In one embodiment, the router may then send the query to a resolver on the hub to perform the query resolution. In one embodiment, a provider may be selected to receive the query if the queryspace specified in the query matches a queryspace of the provider and the path predicate specified in the registration message selects a non-empty set of nodes in the query. In one embodiment, the resolver may index and search provider information for queryspaces that match the query.

After determining the one or more providers to receive the query, the hub may route the query to the one or more providers as indicated at 304. In one embodiment, the resolver may provide a list of the selected one or more providers to the router. The router may then send the query to each of the selected one or more providers. Once a provider receives the query, it may search for results in its queryspace that satisfy the query as indicated at 306. A backend search engine of the provider may perform the search. In one embodiment, the query may be translated from the query routing protocol to a protocol used by the provider by a query routing protocol interface of the provider. In one embodiment, a provider QRP interface or adapter may access a backend search engine of the provider to perform the search.

The provider may compose a response (containing the results of the query) and send it back to the hub as indicated at 308. In one embodiment, the query response may be translated from the protocol used by the provider to the query routing protocol by a query routing protocol interface or adapter of the provider before sending the response to the hub. In one embodiment, the response may be received on the hub by the router. The hub may receive one or more responses from each provider that was sent the query at 304. As indicated at 310, in one embodiment, the hub may collate the responses received from the one or more providers prior to sending them to the consumer. The hub may be configured to tailor the collated responses, as by arranging them in a particular order or according to some categories, by chronological order, to indicate relevancy, or some other method that may be useful to the consumer. The hub may then forward the (possibly collated) responses to the consumer as indicated at 312, and thus to the querying application. In one embodiment, the router handles the routing of the response(s) to the consumer. The consumer may receive the query response and optionally display the results as indicated at 314. Optionally, the consumer can do whatever is necessary to the results, including storing the results, forwarding the results, and modifying the results. In one embodiment, the query routing protocol interface of the consumer may translate the query response from the query routing protocol to a protocol understood by the consumer after receiving the response from the hub. In one embodiment a consumer QRP interface at the hub or a consumer proxy may translate the query response from the query routing protocol to the protocol understood by the consumer.

In one embodiment, instead of, or optionally as well as, sending the results to the hub, the provider may send the results directly to a location specified in the query message. For example, the query message may specify a URL that the consumer wishes the results forwarded to or displayed at. As another example, the query message may include an email address or addresses that the consumer wants the results emailed to.

In some embodiments, pre-crawling may be employed to create or update a provider registration automatically. For example, a provider may register with a distributed information discovery network. The provider may use, or contract a service to use, a tool to build a statistical metadata index from documents retrieved automatically through the provider's web-based interface. The metadata index may then be used to provide query routing. In other words, the provider's site may be "crawled" to create the registration (e.g. an XML-based metadata index). Key terms may be selected as the site is crawled to form the registration index.

A queryspace is a unique identifier for an abstract space over which a query will travel. Queryspaces may be identified by unique URIs. Queryspace URIs may not necessarily reference actual content. Queryspace URIs are identifiers that providers and consumers may use to find each other. In one embodiment, both providers and queries may have queryspaces. A provider's queryspace may be defined as a schema that defines the scope of the set of data which the provider is capable of searching. A query's queryspace may be defined as a schema that defines the scope of the set of data which the consumer wishes to search.

In one embodiment, the distributed information discovery platform may not make assumptions about the syntax or semantics of queryspaces. In this embodiment, the distributed information discovery platform does not process queryspaces, nor does it attempt to validate queries and responses—queryspaces are purely for coordination between consumers and providers. In one embodiment, a queryspace may include information regarding structure, for example so that queryspaces may allow providers and consumers to agree on the structure of messages and by specifying structural constraints in a standard form, e.g. a DTD or an XML Schema. In one embodiment, a queryspace may include information regarding semantics, for example so that providers and consumers may agree on the meaning of the messages that they exchange (in addition to their structure). While structural information may be machine-readable, semantic information may be intended for use by in writing client and server software. In one embodiment, a queryspace may include information regarding ranking. Queryspaces may define how clients may sort the results that they receive. Ranking may be application-dependent, and some applications may not require ranking at all.

In one embodiment, the distributed information discovery platform may not specify methods for exchanging queryspace information. The distributed information discovery platform may ensure that providers receive only queries that match their queryspaces. The distributed information discovery platform encourages efficiency by allowing providers to filter the queries that they receive. To filter queries, a provider may include one or more predicates with each queryspace that they register. A predicate statement may be applied to each candidate query in the given queryspace; only queries that match the predicate statement may be sent to the provider. Internally, the distributed information discovery platform may use the predicates to optimize routing.

In one embodiment, each query may contain at least one query section which may contain arbitrary XML. The contained XML should conform to the specified queryspace; otherwise, the query will probably not match any information provider predicates and will therefore receive no responses. In some embodiments, the distributed information discovery platform may not attempt to validate the query. If multiple query sections are specified, the information provider may choose which query to respond to. In one embodiment any QRP interface may indicate that a query cannot be processed, for example if it is an illegal query or otherwise invalid. In one embodiment, a resolver may validate a query according to a registered schema for the queryspace identified in the query.

In one embodiment, the query routing protocol does not require queries or responses to identify machine addresses. Some queryspaces may agree to share addresses explicitly (e.g. peer-to-peer file sharing), while other queryspaces may choose to share addresses implicitly (e.g. with embedded XHTML). The structure of both the query and the response may be specified (explicitly or implicitly) by the chosen queryspace. In an example of a full-text schema, the response in the data section may be mixed-content XHTML to be displayed in a browser. In an example of a music schema, the data section of a response may contain structured information intended for applications as well as "unstructured" XHTML intended for humans.

Some embodiments may use full-text queryspaces. In one embodiment, a full-text queryspace may use the following DTD:


	<!DOCTYPE query [
	<!ELEMENT query -- (text?)>
	<!ELEMENT text -- (#PCDATA)>
	]>

For example, a query for "dog biscuits" under this queryspace may be formatted as:


	<query>
	<text>dog biscuits</text>
	</query>

In one embodiment, a full-text queryspace may be the default queryspace. In some embodiments, a full-text queryspace, such as the above example, may be extended to support "and" and "or" operations.

Providers may register query predicates with a distributed information discovery network, e.g. by registering with a hub. When a client submits a query to the network, it is resolved to matching providers. For example, a provider may register a registration using the queryspace specified by the URI "http://www.infrasearch.com/food/recipies":

<queryspace>http://www.abcd.com/food/recipes</queryspace>

<query-server>http://www.efgh.com/search.jsp</query-server>

<predicate>baba ghannouj ghannoush ganoush</predicate>

<and>

<type>appetizer</type>

<ingredients>eggplant tahini</ingredients>

</and>

</predicate>

</register>

This registration registers the provider with the recipes queryspace with two predicates. Queries with "appetizer" in their <type> node and either of the words "eggplant" or "tahini" in their <ingredients> node are matched by this registration. A predicate is also registered that will direct any query containing any of the words "baba", "ghannouj", "ghannoush" or "ganoush" to the provider's query server running at http://www.efgh.com/search.jsp.

Query Node Patterns (QNPs) may be the basic building block of query predicates. Each matches a node of an XML query. QNPs may be XML fragments. They match a query when they match some subset of that query's structure, or, more formally, they may be constructed by a series of the following transformations: (1) deleting a node in the query; or (2) replacing the query with a subnode of itself.

For example, consider the following XML query:


	<request>
	<object type=file>
	<format>mp3</format>
	<artist>U2 Nirvana</artist>
	</object>
	</request>

This query is matched by the QNPs as illustrated in Table 1 of FIG. 9.

In QNP matching, tag text (a.k.a. character data) may be tokenized at whitespace breaks and considered a set of tokens. Some embodiments may be limited to keyword matching only. Other embodiments may support phrase matching as well. In some embodiments, matching may be case-insensitive.

In some embodiments, a QNP may only contain one path through the query XML. In such embodiments, the following QNP would be invalid:


	<object>
	<format>mp3</format>
	<artist>U2</artist>
	</object>

In other embodiments, the single path restriction does not apply and the above QNP would be valid. In single path restricted embodiments, the above QNP would instead be specified as a predicate containing the conjunction of two separate QNPs:


	<and>
	<object>
	<format>mp3</format>
	</object>
	<object>

	<artist>U2</artist>
	</object>
	</and>

Tag text may be an exception to the single path restriction. In some embodiments, if a QNP node contains multiple text tokens, these may form an implicit disjunction.

A query predicate may be a boolean expression composed of QNPs. In some embodiments, predicates must be in conjunctive normal form, i.e., a conjunction of disjunctions. In other embodiments, this restriction may not apply.

As an example of a conjunctive normal form predicate, consider the following query predicate:


	<predicate>
	<and>
	<object type=file>
	<object><format>mp3</format></object>
	<or>
	<artist>U2</artist>
	<artist>Nirvana</artist>
	</or>
	</and>
	</predicate>

Note that the first two conjuncts are implicit disjunctions. When an <or> . . . </or > tag contains only a single QNP, the <or> . . . </or> may be dropped. Similarly, if the top-level only has one element, the <and> . . . <and> may also be dropped. Thus, according to one embodiment, at its simplest, a predicate may be of the form:

This predicate would match any query containing the word "U2" or the word "Nirvana."

As mentioned previously, a resolver may create and maintain a set of indices for the provider registration files, with separate indexes for each queryspace. When a provider sends a registration file, the resolver parses it into a set of predicates, each predicate having a set of clauses, and each clause having a set of disjunctions. In one embodiment, predicates may be in conjunctive normal form. Each predicate may be given a global unique predicate ID, and each clause may be given a local clause ID. For each pattern in the registration, a posting may be created which contains the predicate ID and the clause ID. The predicate ID and clause ID may be used to trace the pattern to the clause in the registration where the pattern occurs. The (pattern, posting) pair may be stored in the corresponding query space index. The posting may also include a score, which may be updated based on feedback received from the user. The following is an example of a simple XML fragment of two predicates from a registration and the corresponding index entries:


	<predicate>
	<and>
	<object type=music>
	<object><format>mp3</format></object>
	<or>
	<artist>U2</artist>
	<artist>Nirvana</artist>
	</or>
	</and>
	</predicate>
	<predicate>
	<and>
	<object type=movies>
	<object><format>mpeg</format></object>
	<or>
	<title><quote>Little Mermaid</quote></title>
	<title><quote>Snow White</quote></title>
	</or>
	</and>
	</predicate>

The corresponding entries in the index may be:


	object&type=music	(predicate0, clause0)
	object>format>mp3	(predicate0, clause1)
	artist>U2	(predicate0, clause2)
	artist>Nirvana	(predicate0, clause2)
	object&type=movies	(predicate1, clause0)
	object>format>mpeg	(predicate1, clause1)
	title>Little Mermaid	(predicate1, clause2)
	title>Snow White	(predicate1, clause2).

That is, the index will have eight entries. In one embodiment, at least three of these entries have to match a query for the query to be routed to the provider.

Query resolution is the process of determining a set of one or more providers to which a given query should be routed. Sending all queries to all providers is inefficient, therefore the distributed information discovery platform defines a framework for providers to register the type of queries they are interested in receiving and provides a query resolution and routing service. Providers may specify the type of queries they wish to receive in their registration file.

In one embodiment, the minimal condition for matching a query to a provider is that the query has to have the same queryspace as the provider registration. In some embodiments, the minimal condition for matching a query may be for the query to have at least one matching element to the queryspace of the provider registration. In one embodiment, the set of providers may be selected by the resolver 102 in a certain order. In one embodiment, providers which have all clauses of at least one predicate satisfied may be selected first. In order to match a predicate, a query may first be tokenized into a set of patterns (QNPs). In one embodiment, providers may be ranked based on the matched pattern scores. In one embodiment, providers which do not have a matching predicate, but are similar in their responses and have the same queryspace as providers who have a matching predicate may be selected in a lesser category. In one embodiment if the number of providers returned is still less than the maximum, a provider may be selected (e.g. at random) from the same queryspace as the query. In this embodiment, this allows the exploration of the provider content in case the provider registration file is incomplete, or is not updated frequently.

As mentioned previously, there may be a score associated with each (pattern, posting) pair in the resolver index. In one embodiment, scoring may be used to determine the popularity of providers for a particular type of query. Scoring may be used in selecting the most popular providers relevant to the query first. Scoring works as follows. If a user sends some feedback in response to a query response, the (pattern, posting) pairs that matched the query may be retrieved from the corresponding queryspace index, and their scores updated (i.e. increased for a positive feedback or decreased for a negative one). In one embodiment, a simple score update formula may be used:

where (0<alpha<1) determines the rate of change of the score. Other embodiments may use other score update formulas.

In one embodiment, in instances where there are very few providers who match a query, providers may be selected that did not match the query, but who have registered the same query space and who are similar to a provider who matched the query. In one embodiment, a method similar to collaborative filtering may be used in determining provider similarity. Providers who tend to match the same queries are considered more similar. In one embodiment, a similarity matrix may be maintained in the resolver. The entries in this matrix may determine the degree of similarity between provider x and provider y.

A router may perform the certain functions. For example, in one embodiment a router may receive the queries from the end application/consumer. In one embodiment a router may route the queries to the appropriate providers. In one embodiment a router may merge the results of the queries and presents them to the end application. In one embodiment, a router may include routing or address information with its communications.

When the router receives a request from the network, it may ask the resolver for a list of nodes on the network that are registered as wanting to receive queries like the request received. Once the resolver returns a set of network node endpoints, the router routes the query to this set of providers. In one embodiment, the resolver may return network node IDs with the network node endpoints that may be relevant only within the distributed information discovery platform and that may be used for logging.

In one embodiment, a router may be a JAVA Servlet. The router may be platform-independent so that the deployment platform for the router may be Linux, Win32, etc. In some embodiments, routers may be distributed or clustered.

In one embodiment, a router system may be organized to include a router to perform certain functions, for example functions described above. In one embodiment a router system may include a RouterServlet to receive routing requests and give access to real-time statistics. In one embodiment, a router system may include a HttpRouteConnection to use HTTP as transport and XML as encoding for a route to a given provider. In one embodiment, a router system may include a Router.Stat to provide statistics for a given route, for example bandwidth, response times, traffic, etc.

The RouterServlet may receive a request to route a particular query. In one embodiment, each routing request may be an HTTP request with certain headers. For example, a uuid or unique identifier for the request (which may be used for logging purposes in the router, and may have other uses in other components or users of a distributed information discovery network). Another header may be a timeout or the amount of time to give each provider to respond. In one embodiment another header may be a NumHits, where each provider may respond with several hits but the router may take only the first N hits to be propagated back to the app/user. Another header may be a FlushAfter that may indicate to flush the response stream after receiving responses from N providers.

In one embodiment, the body of the routing request may be an XML-encoded query (see description of queries above). In some embodiments, the routing request may also include a set of cookie headers, which may be encoded, for example, as

When RouterServlet receives a query request from the distributed information discovery network, it asks the resolver for a list of nodes on the network that are registered as wanting to receive queries like this one. The resolver may return a set of network node URLs and network node IDs (e.g. unique provider IDs stored within the distributed information discovery router system and used for logging). The Router may then route the query to this set of providers.

The router may contact the list of providers returned by the resolver. At least one QRP interface may be used when the router contacts the list of providers. In some embodiments a router is not limited to any transport or encoding scheme. In one embodiment, different transports and encodings may be plugged in. In one embodiment, HTTP and light-weight XML encoding may be used.

In one embodiment, the router may use an exponential back-off algorithm to handle spamming and/or slow or temporarily down hosts. For example, if a provider exceeds a set timeout, the resolver subsystem may be alerted to make the provider no longer active in the subsystem. If a time-out is exceeded, or exceeded too often, the provider may be unregistered or flagged so that further resolutions do not include this provider.

In some embodiments, in addition to collating the responses from the providers, the router may also pass through HTTP cookies (cookies may be retrieved from and set on a URLConnection class via a get/setHeader method, so this may be transport-independent, since other network-transport implementations of the URLConnection interface may be used). When passing a cookie from a provider to the client, the router may encode them as "unique_provider_id=base64encoding_of real_cookie", for example, so that it may later match cookies with provider IDs when the user does another search.

In one embodiment, the Router may receive a query in XML format through a HTTP interface. When the Router receives the query, it may sends the query to the Resolver through an HTTP Interface. The Resolver may return with a list of providers that have registered interest in this query. In one embodiment, the Router does not attempt to interpret the query at all. The Router may then set up multiple threads, each thread opening a URL to post the query to each of the provider. In one embodiment, the query may be posted to each provider with a timeout value. When the provider returns with a result page (e.g. in XML), the router may parse the result page and extract the "hits" to be merged with the other hits from the other providers. The number of hits, the timeout value and the number of provider results may be specified through a "preference" interface.

In one embodiment, the router may maintain a pool of TCP/IP connections to the providers and reuse them. This reuse may reduce the overhead in opening and closing connections. For example, each HTTP request to the providers may use KEEP_ALIVE so that the connections will not be closed by the provider.

In one embodiment, the router system may track certain statistics so that administrators may access the router system to view current statistics about their node, such as how many queries were sent to them today, what's the average response time, how many queries failed, etc.

A provider may be registered with multiple distributed information discovery routers. In such embodiments, real-time stats may be aggregated at the time of viewing by code in the provider subsystem. This code may query each router to give up-to-the-moment stats for a given provider. The resulting information is processed and displayed.

In some embodiments, a distributed information discovery router may store and allow administrators to view historical data about their nodes. In an embodiment, each router system may keep a local log of its actions and export the log for download via HTTP with authentication protection. In one embodiment, logs may be periodically aggregated to a log-administrator machine with the script. Once aggregated from all the router systems, the logs may then be parsed. The result of the parsing may be a set of logs per provider.

In one embodiment, each parsed set of logs may include a log file, for example with information regarding the router noted down for that provider's ID (e.g., provider-id.log). In one embodiment, each parsed set of logs may include information regarding successful routes of requests for that provider (e.g., provider-id-success.log). In one embodiment, each parsed set of logs may include information regarding failed routes of requests for that provider (e.g., provider-id-error.log).

In the above example, the log file may be available for download by the administrator, so that a human administrator may run his own set of scripts on that data and maybe glean something only he wants to see from it. A log may be plotted for each provider (e.g. using gnuplot during the parsing), so that the provider-human, who doesn't know how or doesn't have the time to pipe the log to his own charting tools, may visualize the correspondence between time, number of successful routes, and number of failed routes. In one embodiment, a log file or parts of a log file may be accessible to applications or elements of the information discovery network.

In the above example, a failed route may be one where the provider didn't accept the connection or took