Tim Bray wrote:
On Mar 30, 2006, at 9:20 PM, James M Snell wrote:
I would agree that, as a best practice, the xml:base should appear on
the content element, but implementations need to be prepared to use
whatever the in-scope URI is (e.g. if no xml:base is specified, relative
refs in the content will be relative to Content-Location or the feeds
Request URI).
Maybe. Highly error-prone.
Not sure what you mean by highly error-prone, but I do know that support for
Content-Location in aggregators is essentially non-existent. I've run tests
on 16 different aggregators and Snarfer was the only one that supported
Content-Location as a base URI. Thunderbird was the next best in that it
made use of the Location header when there was a redirect. A couple of
others at least used the request URI. However the rest either used the feed
alternate link, the element alternate link or the server hostname. Two
didn't seem to support relative URIs at all.
Aggregators tested: Blogbridge, Bloglines, BottomFeeder, FeedDemon,
FeedReader, Google Reader, GreatNews, JetBrains Omea, Netvibes, Newsgator
Online, NewzCrawler, RSSBandit, RSSOwl, Sharpreader, Snarfer, and
Thunderbird.
I also understand there is some debate whether supporting Content-Location
is a good idea at all (at least in web browsers). Firefox at one point
started adding support, but they determined that it caused problems with
broken servers (mostly IIS I believe). I know there was some discussion
about doing server detection and working around those servers, but some
other issue came up that made them give up the whole idea (I can't remember
what). I'm not sure whether any of these issues apply to feed readers.
Regards
James
