- From: Frank Carvalho <dko4342@vip.cybercity.dk>
- Date: Wed, 05 Oct 2011 00:01:20 +0200
- To: Giovanni Tummarello <giovanni.tummarello@deri.org>
- Cc: semantic-web@w3.org
Hello Giovanni When we started the SOA programme some 7 years ago, the first project was to establish a full technical framework for the SOA. This included a dedicated metadatabase with the express intent to cover all bases with respect to the entire SOA, including specifications, documentation, integration to live environments etc. etc. And a specific state-of-the-art product was chosen. In parallel to this work, I began developing the skeleton for the fully automated RDF setup that we have now. After 5 years of development on the proprietary metadatabase it was abandoned, in favour of our own setup. The reason is that our own setup was, and is, vastly superior to the proprietary system. The proprietary system ran into many problems dealing with versioning. It simply didn't scale, and the way it was designed had no hope of every working. Also the import/export mechanisms were way too obscure to be useful for the average IT supplier. But by far the biggest problem was the lack of flexibility in defining the metadata model of the metadatabase itself. All operations on the database relied on a very strict metadata model, and everytime this structure was challenged by real life needs, it would require expensive changes to the metadata model. In contrast, our own homegrown setup had a simple and efficient strategy for versioning, the ultimate open-ended metadata model, tons of (RDF) tools available and last and not lest, way better performance that the proprietary solution. Even when we ran it in an XML-database! I have of course been curious every time a company boasts the solution to end them all in metadata management, but so far I have seen nothing with the kind of flexibility that we can provide ourselves. I have noticed that everybody still promotes products with fixed metadata models, and as long as this is the case, I am not buying it. So will I still defend the choice we made? You bet! Best /Frank > Frank thanks a lot for the account, it is very rare at least for me to > heard such interesting report dealing with actual deployments in > production > > I know you've been around the list and so possily in love with the > technology, so a devil's advocate question can be : do you feel > confidently that you've explored alternatives to custom rdf > development? e.g. have you explored the existing landscalpe of soa > governance products before starting this custom effort ? are you aware > of alternatives out there and would you still defend this choice? > > not hinting at anything :) just sincerely curious. > thanks again. > Gio
Received on Tuesday, 4 October 2011 22:01:47 UTC