Re: bug with gitweb on kernel.org

On Sun, 2007-04-22 at 21:16 -0400, Nicolas Pitre wrote:

On Sun, 22 Apr 2007, J.H. wrote:

quoted

On Thu, 2007-04-19 at 23:02 -0400, Nicolas Pitre wrote:

quoted

Almost 2 months ago we discussed about gitweb not properly detecting the 
client's ability to deal with application/xhtml+xml, something to do 
with the caching of a previous request from a client which did support 
it and serving the same content to a subsequent client which does not.

I apparently missed that entire conversation, my apologies.

quoted

Right now www.kernel.org/git is unusable for me with lynx as it keeps 
prompting:

	application/xhtml+xml  D)ownload, or C)ancel

Is there any plan to have that fixed?

Well there are a couple of quick thoughts, so far (in my quick testing)
lynx and IE are the only two browsers that have issues with this
particular bit of code.  Links, konqueror, safari, firefox, mozilla, etc
all seem to handle the pages without issue.

No.  You also missed that links, elinks, and the emacs one (w3m or the 
like) were also reported to fail.  And sometimes lynx even works.

A more detailed set of tests (cache file included as attachment for
those curious, also the cache file was set as read only after it was
generated by epiphany so it was not regenerated during testing):

- epiphany v.2.16.3		- passed
- elinks v.0.11.1-5.1		- passed
- elinks v.0.4.2		- failed
- lynx v.2.8.5rel.1		- failed
- lynx v.2.8.5dev.7		- failed
- konqueror v.3.5.6-0.3.fc6	- passed
- konqueror v.3.1-13 Red Hat	- passed
- I.E. v.6.0.3790.1830 (2k3)	- passed
- I.E. v.6.0.2900.2180.xpsp_sp2_gdr.070227-2254	- failed
- w3m v.0.5.1			- failed
- w3m v.0.3.2.2			- failed
- firefox v.1.5.0.10		- passed
- Seamonkey v.1.8.1.2pre	- passed
- Mozilla v.1.7.13		- passed
- Mozilla v.1.4.2		- passed
- safari v.1.0.3		- 50/50 - loads but doesn't render anything after
description link (viewing source shows missing content)

Thats everything I have immediate access to (installed and up and
running) - if we need more data points I can start installing older
versions of netscape and even backtrack through several version of IE
(win95,win98,winME,win2k) but my guess is that most of the gecko based
platforms that are even remotely recent are fine, same goes for
konqueror.  I.E. is horked on XP until I.E. 7, win2k3 has a fix for this
already.  eLinks seems to be handling it with something relatively
recent.  And more or less everything else isn't handling the mime type.

quoted

 Taking a quick glance at the code it seems IE claims to be xhtml+xml 
compliant but apparently isn't really (any real surprise?) and lynx 
just doesn't seem to support that mime type.

Lynx and many others.  It is just a question of luch whether the served 
page is acceptable or not.

Well the only difference in the pages being served is the mime type
application/html vs. application/xhtml+xml.  Does anyone know the
original impetus to using application/xhtml+xml (despite the fact that
it's technically the correct choice) vs. just using application/html for
everything?  I'm sure there was a good reason behind it and I'd rather
know what that reason was before I got changing things

quoted

The simplest fix would be to eliminate the distinction between
applicatoin/xhtml+xml and application/html in the gitweb code (or at
least in the caching gitweb code) and have everything claim a mimetype
of application/html and let the browser sort out if it's using xhtml or
html from the doctype.  This would solve both the problem your seeing on
lynx and would make the caching gitweb usable by more IE users.

Great.

I'm going to wait for further comment, if I don't see anything by the
end of Monday (PDT) I'll go ahead and make the change and get it out to
kernel.org.

- John