Let’s be clear, semantic web technologies, particularly OWL and RDF, are not bad things. I’ve seen some great applications put together using these in ways that would have been very hard to achieve with say more traditional relational database approaches. They’re probably a near perfect fit for governance repositories (SOA or other), which was my argument all the way back in 2004 when Hugh and I first presented our work on SOA governance. I just don’t think they’re a panacea for whatever ails financial services (or even financial services IT).

The trouble is that semantic web technology isn’t as accessible as regular web technology. That’s why it takes people like you, with substantial intellect and years of experience, to work with the raw technology and the platforms using it. I would wager that even with your background learning how to use RDF and OWL in an impactful way was a task on a par with learning a new programming language.

I dearly wish that TimBL’s vision could come true – that we could share meaning by making a little more effort; but I feel that’s like wishing for a universal human language, or world peace (do you get one without the other?). Perhaps if more people were learning how to use these tools things would be easier, and maybe that’s the point of the web science brigade. At the moment though it seems that you have a problem if you need somebody that understands derivatives OR XML schema, and a real headache if you want somebody that understands derivatives AND XML schema.

PS UDDI has in my view always been undead – a zombie technology stumbling around causing trouble. It was never really fit for the purpose it was designed for, never mind the purposes people subsequently tried to turn it to. Of course its worst failing was the use of t-models, which people found obscure and hard to understand, which kind of brings us back full circle.

PPS ATM’s post proclaiming the death of SOA was deliberately provocative (and so by implication was my link to it). Most enterprises haven’t gone to the trouble of establishing the right culture to allow SOA to succeed. If you find yourself at a place where the incentives are correctly aligned to support what our mutual friend Joe calls ‘reuse in the large’ then that’s a lucky break.

As you have said in the past there is a lack of maturity in the tools out there to really make SOA and in particular the semantics easy for the masses. So rather than proclaim “SOA is dead” some have been busy creating a workable solution – a pragmatic attempt to resolve the issues you mention of “shared meaning”, “ontology” and the like.

The solution that has been implemented utilizes an IBM product, Rational Asset Manager (RAM) which provides us with a flexible meta data repository that allows for assets (can be a server, a service or a document) to be classified against an organizational taxonomy, functional domains, vendor, technology and other taxonomies – which is fine, but the power comes in the ability to create relationships between assets. As assets are pumped into RAM via the API’s they are classified, and relationships created to other asset types, fed by/feeds, consumes/consumed, contains/contained by etc….

So what does this look like, well after every application, server, database along with operational data, eg, outages, trouble tickets, source code metrics, runtime metrics and so on from the myriad of different sources are added. The “assets” take on a shared meaning such that a question like “which applications had more then 10 outages in January that runs on Solaris 10.x and uses Oracle 9,x and belongs to the Asset Management group X and consumes the Java logging service?”

In the SOA world, reusable assets submitted to RAM also take on shared meaning in their implementation, deployment, usage from a consumer/provider as well as fed by/feeds…..

The current implementation of my RAM Repository has over 20k of assets

You mention Governance in similar vain, well with IBM integration with WSRR (WebSphere Registry Repository that uses OWL for it’s classification btw) along with Amberpoint SMS we are able to demonstrate closed loop governance – but I should save that story for a blog of mine own

As far as anything that should be dead – UDDI should be.

I should add that having such a repository with an ability to create assets, classifications, relationship types and capabilities without being a “web science sycophant” we are also able to model reference architectures in RAM….. use modeling tools to import/export data….. another story to be told another day, but then I would have to find a blog for pragmatics

Thanks for blogging – the blogs are quite interesting and I should comment on more of them :-)

Regards

Thomas Wood
Persona non Grata