- From: Paul Prescod <paul@prescod.net>
- Date: Mon, 25 Nov 2002 15:38:52 -0800
- To: Anne Thomas Manes <anne@manes.net>
- CC: www-ws-arch@w3.org
Anne Thomas Manes wrote: > Well for one thing, an http: URI lends the impression that you might > be able to perform an HTTP GET to get the resource identified by the > URI. But you can't do an HTTP GET on, say for example, a tModelKey. If you could do an HTTP GET on a tModelKey that would imply that every organization owned their own tModelKeys and that the system had no single point of failure. If you "discover" a tModelKey in some data, you could resolve it without trying to determine WHICH of the UDDI servers might be able to give you information about that UDDI. In other words, UDDI as it exist is designed to be a centralized system with hierarchical control when it should be a decentralized system where everybody controls their own information and anybody can become an aggregator of the information. examples of Web data aggregators include AltaVista, Google, Meerkat and Yahoo. So I'm not against registries/repositories. But they should The centralized model is architecturally unsound. And as far as I know, has no advantages whatsoever. (I'm open to arguments on that point, however) I'm sorry to venture towards conspiracy theories but I have to say that as far as I can see, the only reason someone would come up with such an architecture is because in the back of their mind they see UDDI as being analogous to DNS and they wish to be the NetSol or ICANN of UDDI. This will not work because it will not scale as well as the decentralized solution. > Currently, the only way to get the resource identified by a uddi: > scheme is to issue the proper UDDI Inquiry API (e.g., get_tModelDetails). Not only issue the proper query, but to the proper UDDI server (or one in its federation. Given only a UUID or UDDI URI, you cannot know whether to use it inside the firewall or outside. (and which inside or outside). > At some point in the future (maybe UDDI V4) the API might be extended > to allow you to do something like > a UDDI GET. IMHO, moving to a decentralized (not federated, but truly decentralized) model should be the highest priority of the UDDI V4 project. But it is your standard, not mine. Paul Prescod
Received on Monday, 25 November 2002 18:39:41 UTC