James Governor's Monkchips

Any viable service registry should have UDDI as its foundation.

Share via Twitter Share via Facebook Share via Linkedin Share via Reddit

I found the above statement in a blog post by Michael Curry, of Ascential (now IBM).

His new view was based on some pushback by Systinet to another post.

Hats off to Systinet for its aggressive advocacy of UDDI as a foundational technology, associating itself strongly with the standard, which is good marketing practice, even though, in the words of Tom Erickson:

Most of our customers are certainly looking for capabilities beyond UDDI, thus our development efforts to create Systinet Registy 5.0 a year ago and 6.0 which we announced last week. Our new platform Blizzard, available in early access next month, takes things a step further by adding a repository and capabilities to model the SOA.

I better not mention ebXML. But then again i better not mention LDAP either. Or DNS. Or JNDI. or bloglines, for that matter.

I think i probably need to talk directly to Systinet, because I clearly don’t get it. Or Michael doesn’t. Or someone doesn’t.

I don’t believe in total homogeneity (even of standards), because i have never seen it, and where I do see tendency towards it the results are often unhealthy [monoculture and boll weevils], but its hard to argue with DNS as a foundation for the Interweb. Maybe there is a different set of factors at work in the world of SOA that is different from business as usual.

Amazon apparently agrees WSDL is hard, and if WSDL is too hard, what does that say to UDDI?

Tim Bray may be a member of the loyal opposition, but i am certainly not going to discount his arguments just because he is.

I spoke to Duane Nickull at JavaOne last week. The guy is ungodly smart. He is thinking big and this presentation is worth checking out for the kinky front page [on that note i met a guy on my flight home from JavaOne, Bill Ennis, who owns Lara Croft. Well, his company does anyway], let alone the rest of the content. Duane has a very big IDEA (intelligence driven enterprise architectures). [Note to Duane: “enterprise” is so last century… how about IDSA (intelligence driven service architecture. not as catchy i know but “enterprise” is surely too normative about business entity granularity)]

Duane says we should not define SOA by referencing an implementation. I agree.

Dare Obesanjo is thinking clearly about service orientation issues by questioning whether

1. We need a standard service description language at all

2. If we did have one would it have to be XML-based, anyway?

I also had a really good hour at JavaOne last week with Miko Matsumura of Infravio. Funnily enough for a shadow boxer he had the great idea of doing the meeting upstairs in the sunshine, so i got a sunburned nose. We had a far better conversation because of the fresh air and sound of running water, and i was able to put forward my model for Architecture as a Service (AAAS)… 😉

Anyhow while we were talking I put forward a heretical, linguistic analytical position. When the folks who first put forward UDDI and WSDL were wondering how to articulate their concepts to the market they thought yeah Yellow Pages, that ouught to do it. Unfortunately over time the metaphor became the reality. Rather than using Yellow Pages to explain different concepts they began to use it as a design point, which has led to some strange effects.

Sorry to wax philosophical but we need to be capable of abstraction to contextualize all of this stuff. Wittgenstein put forward some powerful arguments about the mistakes we make when we confuse reference with implementation. We bump our heads against the limits of the language. The meaning of a word is its use in the language. This is natural extensibility.

There is no directory of language we can refer to, no single implementation. Dictionaries can help with language. That is not to say they are language. Like ITIL they are descriptive about best practices. Grammar books are sets of rules (that are made to be broken). For some great writing on the subject of language standardization I would heartily recommend The Stories of English by David Crystal [available now for $14]. Maybe its me getting confused with the metaphor (human language as computer language).

Yellow Pages as a mechanism for managing every service we could ever need or describe…

One problem? See the sizing limitations in the Microsoft UDDI implementation, for example.

Limits to the number of business entities and service specifications have been established as defaults for publisher accounts. By default, an information publisher may register only one business entity, four business services and ten service specifications within the registry. Requests to have this restriction removed can be made by sending email to [email protected].

This is not a criticism of Microsoft’s UDDI implementation so much as argument about registry scalability and limitations. I have to email someone for an exception, see. Business in real life is more peer to peer than the yellow pages. These days when i get a call from Yell or BT or any other business registry service offering me a placing you know what i say? No thanks. I am capable of listing myself. Those that need to find me can do so through google or whatever.

UDDI and flexibility. What is Google? Is it search engine? A software vendor? A financial services organization? Or all of the above…What about DoCoMo- wireless carrier, media company, or financial services company? It changes every day. Business entity is somewhat fractal.

Nokia?

Descriptions can’t be once and future because things change so fast. 

Some of my views here tack into tagsonomy thinking, standing on the shoulder of the giant Clay Shirky. Ontologies have a place, but they don’t solve all known IT problems.

Will tModel really be a flexible enough approach to extending businessEntity to describe any business entity? Fedex accounts are named individuals not companies. Surely modelling also has a place here. There are some like Michael Earls who argue every major RSS-based service out there should be jumping into UDDI too (note its an advocacy statement rather than one of fact). But when does a service become major, and when does it become important to jumnp aboard?

Then there are questions about identification, trust and risk. I know who you are by what you do. If i ran an analysis across every utterance you ever made the patterns and word usages would describe you just like a fingerprint. Many business relationships work by doing this in implicit fashion. Do we trust this business? The question can’t be answered by pointing to a directory. Risk is not absolute. Counterparty hierarchy helps us understand risk but it can’t prevent catastrophic failures.

Another synapse heavy discussion last week was with Peter Rogers, CEO of <1060> Research. He brain apparently works a bit like mine but he actually gets things done. Resources and pointers offer fruitful approach to service invocation and manipulation, as evidenced by NetKernel. Could we use a Unix scheduler, rather than a printed book, as a metaphor and design point for service invocation and manipulation? What might that look like?

The Abstraction

Every software component on NetKernel is addressed by URI like a Web resource. A component is executed as a result of issuing Web-like[REST] requests. A software component on NetKernel is therefore a simple software service which hides the complexity of its internal implementation.

Applications or higher-order services are built by composing simple services. Service composition can be written in a wide-choice of either procedural or declarative languages. You can think of this as Unix-like application composition and pipelines in a uniform URI-based application context.

Complexity is managed through the URI address space which may be remodelled and extended indefinitely. A complex aggregated service may always be abstracted into one or more higher-level services surfaced through URI interfaces.

I spoke yesterday to IBM about the IBM WebSphere Partner Gateway version 6.0. The new version includes a cool new function. HTTP over XML. Why? Customer demand.

What if John Sowa is right, that “whenever a major organization develops a new system as an official standard for X, the primary result is the widespread adoption of some simpler system as a de facto standard for X”?

UDDI will be useful, but Michael surely goes a little too far, as does anyone that claims UDDI is the only registry foundation.

But then again – what does viable mean anyway? Defining viability is a little like defining obscenity. An awful lot of companies and technologies in this business are “dead”; but then, that’s a subject for another blog.

 

 

 

9 comments

  1. I tend to agree with you James, and I might go further. UDDI seems to be woefully inadequate in meeting customer needs, and I don’t know of many who are using it in a significant way. Having said that, Systinet seems to have as much or more experience than anyone else out there in terms of how registries are being used in the real world.

  2. There has been some interesting discussion on this subject that James and I have been part of at uddi-dev list. The thread begins at:

    http://lists.oasis-open.org/archives/uddi-dev/200506/msg00005.html

    It had a funny climax (if I may say so myself) at:

    http://lists.oasis-open.org/archives/uddi-dev/200506/msg00031.html

    Thanks as always for sharing your interesting analysis James.


    Regards,
    Farrukh Najmi

  3. Has it occurred to you that all these semantic problems can be solved with tagging? And without the baggage associated with structures and hierarchies. Just thought I’d throw it in…

  4. yes dennis i had considered that. thanks for the idea. i gently pointed to it above:

    “Some of my views here tack into tagsonomy thinking, standing on the shoulder of the giant Clay Shirky. Ontologies have a place, but they don’t solve all known IT problems.”

  5. I have to say I’m not totally convinced with the tagging argument – I’ve argued it both ways on my own site – Sig has interesting things to say about this at Thingamy but I still believe you end up with a mash up between tagging and some form of hierarchical repository. I can’t see any other way that works (but I’m not totally techie) and in any event current ‘search’ just doesn’t seem strong enough to do the bizzo.

    Can’t say I’m totally enamoured with UDDI either. It doesn’t seem to have taken off in the way people expected. And in any event, isn’t UDDI meant to be an enabling thing rather than a core thing. Isn’t UDDI just search by another name but predicated on some person’s idea of what an electronic yellow pages ought to look like? Or am I being my usual simplistic self?

  6. Hi Byron, yes Systinet is experienced with UDDI, but Infravio has a lot of experience with UDDI in the context of integrating UDDI registry with JAXR registry repository.

    This integrated approach came directly from our customer requirements and registry repository integration is being validated by the “Blizzard” platform as well as the Sun Registry.

    We completely agree that more is needed than just UDDI (for example, repository), but of course UDDI is needed to ensure interoperability as it is a key piece of the Basic Profile.

    Miko Matsumura
    Infravio

  7. James,

    I stated in January on the ebXML Vision dicussion list that Knowledge Driven Collaboration will begin to supplant WS-* as the way companies are planning, moving and adopting – and the OASIS ebSOA and BCM TC work is definately providing the basis for implementing that. Duane of course chaired the ebSOA TC and is now chairing the SOA RM TC – so no surprise that he also “gets this”.

    DW

  8. Hi James, good post. Please, take a look at my answer.

  9. some notes on registry evolution from capience:
    http://vstory.techmarket.com/1/08/3/25423_3209272/article_7.html

    capience web site is awful though from a contact perspective

Leave a Reply

Your email address will not be published. Required fields are marked *