Talk:Universal Description Discovery and Integration
From Wikipedia, the free encyclopedia
UDDI is a great concept, however, its widespread use is unthinkable. Would you use someone else's web service from the internet just based on their description of it? UDDI must be used under a strict monitoring program to ensure no one exploit this facility, even then, there may be unintentional bugs and the quality of the WS is never ensured. I don't think this concept will work out in the entire world wide web, maybe it can work in big organisations.
Hoping this gets cleaned up in the near future so we can get some technical information relating to UDDI rather than diverse political commentary.
I think that this WIKKI is great. I am particularly interested in finding folks who have stories to tell about actually using a UDDI directory in their businesses and sharing experiences with them. I work for a large financial conglomerate and we are very committed to UDDI, we are looking for the user’s experience in order to understand what the true best practices are.
In essence our use case for UDDI is the management, and governance of a heterogeneous federated implementation of SOA across regions, platforms, business units, and even spanning from our business to our commercial partners, customers, and vendors. We believe that in order to create a non intrusive network of services that depend upon each other but can have no one infrastructure bus for all messages there is a need for standards based way to define, discover, create, edit, and use the metadata about the services and their lifecycles. We believe that UDDI is the only standard that:
- Defines a data model
- Defines a sense of custody
- Defines a validation API
- Defines a Publish and subscribe API so that one component can cache the information in a directory or notify a directory of a metadata change.
I think the page has drifted into some fairly bitter opinions about MS/IBM, with too much perjorative commentary on their membership of OASIS/W3C, the "so-called" standards body WS-I, and the final paragraph doesnt really parse. I think there is a good argument for having a WS-Politics entry that looks at the politics of the WS-*, rather than having opinions scattered across the pages about the many specifications.
[edit] Opinions galore. What about some links and refs to back them up?
There are a lot of opinions expressed in the UDDI article.
I make no assertions to their accuracy, but, I do wonder if they could be seperated from the technical facets of UDDI, and if there are some online references to articles, discussions, etc. that would support them.
I agree. The author(s) clearly have an axe to grind and this is not the place. Conspiracy theories and loaded phrases like "so-called ws-i 'standard'" are inappropriate in an objective description of technology.
Our company has used the metadata model provided by uddi quite successfully to introduce a valuable layer of abstraction. Our services are categorized and conumers discover locations of their needed services through categorical descriptions. It allows us to relocate, combine, separate, and ensure protocol-version backwards compatability.
In the final paragraph the author refers to this as the "relatively simple role" of binding client systems to implementations, when in reality it is a very important part of operational management. Frankly what the article describes as being the original intent of uddi (machine-to-machine powered free market of service consumption between companies) is something you would never include as part of a highly available commercial system.
I am the author of some of the negative bits, so I must apologise if I made the article more controversial. I dont have any citations to hand, but as a committer on Apache Axis and the author of another SOAP stack, I do have a fair amount of experience and involvement in the WS-* stack. I do occasionally know what I am talking about, and am vaguely acquainted with many of the people in the SOAP/WSDL + related protocols.
Now, if this is controversial, then it must be moved into a split between practise (extensible metadata model) and the original MS/IBM/Ariba vision, of a set of high-availability, mirrored UDDI servers providing dynamic binding from clients to service implementations by arbitrary vendors. That hasn't happened. Why not? Because the service providers don't want to be commoditised, to give the money to the B2B facilitators like Ariba. Why would the provider of a card auth service implement someone else's API, letting you walk away from their service at the click of a switch? No! they want to own you, and own you they will. What we have on the internet is a wide set of custom service descriptions (amazon inventory, amazon S3, Google data, Google Maps MS live stuff, etc). All of these thing achieve availability by DNS, or by building into the protocol something where after logging in you get the endpoint reference (EPR)/URL for the rest of the conversation.
As a result, the original high-availability UDDI system, the mirrored servers, are being switched off. instead of admitting that UDDI has failed, IBM and the others are claiming that this was the intent all the time.
The place where dynamic binding matters today is behind the firewall, where you don't want the expense application hard coded to a server in one location, not if you want to retain the option of moving it later. Now again, you can do a login+redirect in your protocol, or, by using a registry, you get to make the decisions client side, rather than just server side. You also get a single choke point for management tools to instrument and track usage. This is why MS have focused UDDI on the enterprise [[1]]. However, I note that even there a bit of DNS is nice and simple and works with things other than WS-*.
Good point about the history and politics of WS-* needing its own page. That would be a good location for the controversy. For UDDI, we'd have to look at how the business need of the spec authors "create a business where the intermediaries have the control" clashed with the needs of the service providers "own the customers".
Finally, "so called ws-i standard". Who standardises WS-*? It is the strangest standards process ever invented. There is no standards body overseeing it, the way we have for telephones (ITU, GSM), networks (IEEE), or web stuff (W3C). Instead you have vendors defining a spec with the title WS-something, exposing it for review under non-disclosure, then finally announcing it to the world in an already frozen state. [[2]]. All the outside world can do is implement it, use it or ignore it.
SteveLoughran 19:59, 24 July 2006 (UTC)
Most relevant use of UDDI is not in dynamic binding (runtime), but in supporting SOA governance (design-time). UDDI enables interoperability between governance, middleware, development, modeling and whatever other tools consume or produce service metadata. Yet design-time use -- or the fact that UDDI bridges design-time and runtime use -- is not mentioned at all.
[Daniel Feygin] 3 December 2006
