Errant Feed [#354630]

This works in my browser but will not retrieve items. I have tried both CSP and Simple Pie with the same result:

http://www.indystar.com/apps/pbcs.dll/section?Category=TOPSTORIES&templa...

I have seen this before (I think) with a "dll" referenced in the feed url.

If it matters, my server is running CentOS 5.2. The only non-standard software is Lighty (highly recommended btw) instead of Apache.

Any ideas?

Thanks

Comments

Comment #1

gsnedders commented 5 January 2009 at 21:24

The feed is served with two content-type headers, and concatenated together per RFC2616 results in a content-type of "xml, text/xml" — unsurprisingly, this isn't recognized as a feed media type (in the strictest sense it isn't a media type at all). All browsers (basically) treat multiple headers for which only one is expected by ignoring all but the last one (which in this case would result in "text/xml").

Comment #2

eyecon-1 commented 6 January 2009 at 17:59

gsnedders -

If you have a moment I would like to understand how you arrived at that diagnosis so that I can identify these in the future. With almost 500 feeds, one more or less is somewhat irrelevant but I'd like to have better comprehension of the issue.

Thinking about this, I suspect that trying to make FeedAPI compliant - actually non-compliant - would cause a hit to performance. Moreover, I have always opposed open source trying to conform to broken clients/applications. It's a very poor use of development resources.

Comment #3

gsnedders commented 7 January 2009 at 18:39

Using curl -I and looking. :)

As for compliance to specs: the entire web relies upon applications not conforming to specs. HTML and HTTP as deployed have very little similarity to HTML 4.01 and RFC2616 respectively. Ideally, the specs should define how to deal with any byte-stream so that behaviour doesn't need to be reverse-engineered — HTML 5 at least does that. The real issue, in SP's case, is that the entire HTTP client is a very quick off-the-top-of-my-head guess at what will work at the real world, looking at a very small number of documents, and really isn't suited to the sort of design it needs. I also have little interest at the moment in working on it in part because I now work mainly in Python — which has perfectly fine HTTP clients anyway — but also because I have more or less left SimplePie development. Hopefully SP2 will bundle a better HTTP client, and a friend's colleague I know is working on one that should be suitable.