mark nottingham

RSS-Data and Web services

Friday, 3 October 2003

Web Feeds

Jeremy Allaire is writing about something he calls RSS-Data, and I must say it touches on a lot of interesting points. A few;

• data encoding - Jeremy’s view of SOAP Encoding (aka “Section 5 Encoding”) seems to be contrary to the rest of the industry, as evidenced by WS-I’s ruling it out in favour of document-style services, because there is no schema for sec5 encoding. However, I think his problem is more relevant to…
• XML Schema - There’s a lot of pain associated with the complexity and interoperability problems in XML Schema, and some of it shows here. A simpler schema language would be nice, but although Jeremy points out the need, he doesn’t offer a solution, only saying that developers should figure it out in an ad hoc fashion - not a great solution, IMO.
• Asynch vs. sync - it’s well-recognised that async is important; however, there’s a lot of different ways to do it. I suspect that Jeremy’s on to something here; a polling model does give a lot more flexibility, and is much more well-suited to real-world deployments, especially for desktop apps. There are problems with polling, like scalability and messaging lag, but there are ways to mitigate that (we worked on some when I was at Akamai).
• defining a channel - to use RSS as a pub/sub mechanism, we need a much more rigourous idea of what a channel is, what items are in-scope, etc. - see this article for more thoughts.

BTW, I notice that there’s a misconception by some of the folks commenting on Jeremy’s article; XML Schema doesn’t provide semantics, only structure. Anything above that is inferred by users, not provided by Schema.