Nicole wrote:
On 11-Nov-08 My Secret NSA Wiretap Overheard Nicole Saying :
Hello all
I have started to receive complains from people trying to get video's from
msnbc.com that use a # character in the URL.
Such as:
http://www.msnbc.msn.com/id/22425001/vp/27657223#27657223
http://www.msnbc.msn.com/id/22425001/vp/27652443#27652443
The access log shows that it is removing the pound sign and everything after.
7 TCP_MISS:DIRECT
9.2.2.7 - - [11/Nov/2008:09:59:30 -0800] "GET
http://www.msnbc.msn.com/id/22425001/vp/27657223 HTTP/1.1" 200 477
TCP_MISS:DIRECT
9.2.2.7 - - [11/Nov/2008:10:00:18 -0800] "GET
http://www.msnbc.msn.com/id/22425001/vp/27652443 HTTP/1.1" 200 477
TCP_MISS:DIRECT
I cannot see in my config why it would be truncating out the pound
character.
Any assistance greatly appreciated.
On additional i forgot to include:
This seems true for squid 2.6 and 2.7-stable5
cache.log:
2008/11/11 16:33:28| Oversized chunk header on port 59375, url
http://www.msnbc.msn.com/id/3036677
This seems to be true on every browser I test. Enable proxy.. will not load.
Disable proxy (on the browser) and the url loads.
Ah. Bingo.
This is a combination of two problems:
1) the msnbc stream software is sending chunked-encoded response to
Squid when it should not be.
2) and the hack in Squid-2 to cope with that bad behavior has a limit
on the header size it can handle.
You might have to use the Accept-Encoding hack on them:
# Fix broken sites by removing Accept-Encoding header
acl broken dstdomain ...
header_access Accept-Encoding deny broken
PS. an upgrade to 3.1 beta might be an option for you also.
Amos
--
Please be using
Current Stable Squid 2.7.STABLE5 or 3.0.STABLE10
Current Beta Squid 3.1.0.2
