In his post entitled Back in the Saddle Don Box writes

My main takeaway was that it's time to get on board with Atom - Sam is a master cat herder and I for one am ready to join the other kittens. 

This is good news. Anyone who's read my blog probably can discern that I think the ATOM syndication format is a poorly conceived, waste of effort that unnecessarily fragments the website syndication world. On the other hand the ATOM API especially the bits about SOAP enabled clients are a welcome upgrade to the existing landscape of blog posting/editing APIs.

My experiences considerng how to implement the ATOM API in RSS Bandit have highlighted a one or two places where the API seems 'problematic' which actually point more to holes in the XML Web Services architecture than actual problems with the API. The two scenarios that come most readily to mind are

  1. Currently if a user would wants to post a comment to their blog using client software they need to configure all sorts of technical settings such as which API to use, port numbers, end point URLs and a lot more. For example, look at what one has to post to a dasBlog weblog from w.bloggar Ideally, the end user should just be able to point their client at their blog URL (e.g. http://www.25hoursaday.com/weblog) and it figures out the rest.

    The current ATOM specs describe a technique for discovering the web service end points a blog exposes which involves downloading the HTML page and parsing out all the <link> tags. I've disagreed with this approach in the past but the fact is that it does get the job done.

    What this situation has pointed out to me is that there is no generic way to go up to a website and find out what XML Web Service end points it exposes. For example, if you wanted to tell all the publiclly available Web Services provided by Microsoft you'd have to read Aaron Skonnard's A Survey of Publicly Available Web Services at Microsoft instead of somehow discovering this programmatically. Maybe this is what UDDI was designed for?

  2. Different blogs allow different syntax for posting comments. I've lost count of the amount of times I've posted a comment to a blog and wanted to provide a link but couldn't tell whether to just use a naked URL (http://www.example.com) or a hyperlink (<a href=“http://www.example.com“>example link</a>). Being that RSS Bandit has supported the CommentAPI for a while now I've constantly been frustrated by the inability to tell what kind of markup or markup subset the blog allows in comments. A couple of blogs provide formatting rules when one is posting a comment but there really is no programmatic way to discover this.

    Another class of capabilities I'd like to discover dynamically is which features a blog supports. For instance, the ATOM API spec used to have a 'Search facet' which was removed because it seemed to many people thought it'd be onerous to implement. What I'd have preferred would have been for it to be optional then clients could dynamically discover whether the ATOM end point had search capabilities and if so how rich they were.

    The limitation here is that there isn't a generic way to discover and enunciate the fine grained capabilities of an XML Web Service end point. At least not one I am familiar with.

It would be nice to see what someone like Don Box can bring to the table in showing how to architect and implement such a loosely coupled XML Web Service based system on the World Wide Web.


 

Wednesday, 25 February 2004 17:20:54 (GMT Standard Time, UTC+00:00)
WS-IL, the Web Services Inspection Language, was designed to address issue #1, I believe, but nobody wants to talk about it anymore... I actually liked the idea, but the link to the spec seems to have disappeared from MSDN.

http://www-106.ibm.com/developerworks/webservices/library/ws-wsilspec.html
Monday, 05 April 2004 05:19:30 (GMT Daylight Time, UTC+01:00)
http://www.kbcafe.com/iBLOGthere4iM/?guid=20040404201049
Comments are closed.