Tim Appnel proposed using WSIL files to combine RSS auto-discovery and channel rolls into one file format.
Instead of having multiple <link> tags in your index page to point to your RSS files, you could have just one which points to an WSIL file (index.wsil). This file would contain the addresses for all your RSS feeds, but can also contain the addresses for all the feeds in your channel roll/blogroll.
This would make possible something I've been wanting for a while: the 'I've read everything in my blogroll - spread out one degree and show me the feeds they're reading' aggregator function.
The other thing WSIL could be able to contain, as a file format designed for web services discovery, is a collection of links to search engines on other sites. Thus, as I described a while ago, you could start by searching one site, then automatically spread out and search other sites recommended in the WSIL file. If all the search engines returned results in a standard format(s) you could then aggregate all the results together. This would be useful, for example, for following social networks to search weblogs, or using trust networks to search reviews databases (though if I remember correctly the OAI format also allows you to include pointers to related data stores).