Jeremy Zawodny has jump-started the discussion about RSS feed autodiscovery by proposing what he calls RSS Auto-Discovery 2.0 (which I came across via Russell Beattie's RSS and Commercial Aggregators—itself an interesting post). Many good comments follow Jeremy's post as well.

I have been really interested in this concept of "autodiscovery", partially because I think, in some sense, it is like saying that being able to SELECT TIMESTAMP into aVAR in a SQL stored procedure is a way to autodiscover what time it is ;-)

But, actually, what I think is really interesting is that the autodiscovery interaction between a newsreader and a RSS-enabled website seems like distributed robot / agent processing. (I think that Timothy Appnel's RSS Auto-Discovery 2.0 and WSIL, which is his response to Jeremy's post, is a relevant perspective on picturing RSS feeds as web services.)

Jeremy mentions the idea of having a feed description document along the lines of a robots.txt file (which file is used by search engine spiders, see this excellent tutorial for more info), and also mentions variants on the current RSS autodiscovery usage of the HTML link element (see this post for more info).

What I am working on for the iCite net is looking at the autodiscovery interaction and designing it around what, to me, seems to be the thing really going on: interactive querying and retrieving remote processing instructions, between web client and web server program.

In other words, your newsreader is querying a website for RSS feed information, and the webserver is giving the newsreader a processing instruction it can use to find that RSS feed.

I think that interaction also can (and should be) expressed through files that serve a similar role as robots.txt, that is being (search engine or other) spider food. And, the HTML link element is a natural way to indicate a relationship between the content of a web page and other services around that content.

But, with the query / processing concept, it makes sense to me to allow an extensible domain of queries and an extensible domain of results / processing instructions. For example, why not ask if you (i.e., your agent, the newsreader) can have a RSS 1.0 feed with full contents rather than have to figure out if one is offered or accept the first RSS flavor discovered as the only option?

So, a main piece of the iCite net software is the interface for queries / processing of website content. Initially, that interface will function server-side for interactions between web pages / sites. But, my idea is that, if it turns out to be interesting enough, that it will be natural to use that interface in newsreaders, email clients, web browsers, cell phones, etc.

Note: in July, I posted some related ideas about autodiscovery for RSS-type feeds in Maybe RSS autodiscovery link usage should change (in Echo?) and Mark Pilgrim comments on my RSS/Echo/Atom autodiscovery post.