You need to
before you can comment on or make changes to this bug.
I would like people to comment on the attached diff. I think it would be beneficial to add image/svg+xml as a requested image type, to remove the generic XML types (they are still matched by */*), and to make HTML preferred over XHTML (since HTML is better tested).
If a content provider has both HTML and XHTML versions, and they are not equivalent, they can set a qs (source quality) value to boost the XHTML version in negotiation.
Created an attachment (id=12495) [details]
new accept header
Oh, and I removed text/plain too, because that's the apache default for formats it's never heard of, like .backup and .temp, and having it override common things like image/jpeg is bad IMHO.
It would be interesting to know what other browsers (Firefox, Opera and MSIE) send.
(In reply to comment #3)
The current string was copied from Firefox verbatim, bug 9572.
The current string is sent by FF for it's mainResource, but for <img> sources it sends only "image/png,*/*;q=0.5" and for stylesheets "text/css,*/*;q=0.1".
Opera 9 sends "text/html, application/xml;q=0.9, application/xhtml+xml, image/png, image/jpeg, image/gif, image/x-xbitmap, */*;q=0.1" for every request.
IE sends "*/*" for normal requests, and "image/gif, image/x-xbitmap, image/jpeg, image/pjpeg, application/x-shockwave-flash, application/vnd.ms-excel, application/msword, application/vnd.ms-powerpoint, */*" for images.
Most feed readers (desktop & web-based) don't send any Accept string. AppleSyndication sends "*/*".
The Google and Yahoo search bots send "*/*"
The MSN bot sends "text/html, text/plain, text/xml, application/*, Model/vnd.dwf, drawing/x-dwf"
(In reply to comment #5)
> The current string is sent by FF for it's mainResource, but for <img> sources
> it sends only "image/png,*/*;q=0.5" and for stylesheets "text/css,*/*;q=0.1".
Does WebKit send the same string for the main resource and subresources? I believe that would be a regression.
WebKit sends "*/*" for all non-mainResource requests.
This bug has been fixed in r80776 (bug 27267) and I am content with the patch that was applied.