Semantic SOA

Interesting note in Emerging Technologies, re the Semantic Web taking hold in enterprise back ends before the consumer space:

But while all of the pundits were focused on consumer-side Web services (and often predicting the failure of Web services due to the lack of consumer Web services) much of the real work in Web services happened in the nuts and bolts of enterprise back office integration.

Core Web services technologies such as XML, UDDI and WSDL proved invaluable to helping businesses tie together their databases and applications and also integrate with partner applications. And the growth and maturation of these technologies has led to what we today call SOA.

And now SOA is today facing a potentially radical change that could greatly alter the way businesses use and build SOA platforms. And interestingly, the technology that will bring about this change is one that faced a similar growth curve as SOA.

This technology is the Semantic Web, which itself faced misplaced hype in its infancy and is now seeing real growth and interesting implementations that are in many ways different from the initial expectations.

We'd love to say "I told you so", but we didn't - though we did tell some people, but thats because they paid us . Anyway, now its out in the open, lets take the rest of the article:

In short, the Semantic Web is a set of technologies that make it possible to treat all content and information on the Web as if it was data in a database. In a semantic-enabled Web, whole new types of applications and mashups become possible as everything on the Web is tagged, marked and linked to accurately portray what it is.

And therein lies the rub - tagging, markup and linking to Everything on the Web. It is probably mathematically provable that it is an impossible task, never mind getting the tagging coherent and up to date. At any rate, most enterprises struggle to keep even their own databases coherent. The implications of this is that the sunny uplands in the article....:

...in a semantically enabled SOA, the services would actually describe themselves. In these scenarios, it would be possible to write implementations that would automatically find all services and data that had been tagged to meet a specific SOA need (say for example tax information)....

...require a huge amount of pre-work on the data and metadata used.

For this reason we do not think the Semantic Web will break out as auniversal occurrence. Rather, we think it will emerge, like EDI, in specific verticals initially. This is for 2 reasons:

(i) The total "semantic dictionary" in any one vertical is smaller, and I would argue complexity reduces geometrically, not arithmetically - so cost and time to execute are much reduced.

(ii) Some verticals will have decent ROI's, and thus those boats can afford to leave shore earlier, rather than everybody having to leave together in the Great Semantic Webshift.

Net- net, we will probably find that the Semantic Web will break out differently in different verticals, and at some point we will have to map different semantic schemas to integrate them, but realistically that is the way it will probably work out. (In fact there is an argument that says that the Semantic web already exists in small verticals as companies / industries etc agree on the meaning of very limited datasets. In fact, one could argue that "EDI over IP" is an example of the Semantic Web being here already in concept, just unevenly distributed (to paraphrase Mr Gibson)

One of the main drivers of this Splintered Semantic SOA will (we believe) be M2M communications, which by necessity will require very simple transactions, and will probably force a degree of "pidgin protocol" that won't quite be the beautiful flowing contexts that some Sematic Seers forecast.