Re: 1.2beta21+patches Invalid Request

From: David J N Begley <david@dont-contact.us>
Date: Sat, 30 May 1998 14:40:47 +1000 (EST)

On Sat, 30 May 1998, Justin Viiret wrote:

> We're running Squid 1.2beta21 with both patch sets installed under Linux
> 2.0.33. Just thought I'd ask - does anyone else get an error when accessing
> Australia's Altavista mirror under 1.2, at

Timing! This was reported to be by some of our users yesterday and I've
been looking at it today; I *think* I've found a Netscape bug, but would
like others to confirm before I report it to Netscape.

> http://www.altavista.yellowpages.com.au/? We get the following (seems to be
> only the banner advertising - this is not necessarily a problem ( :) ) but
> it's a bit of an eyesore, and it's confusing for the users.):

Agreed - it's only banner advertising on the Australian AltaVista site,
the rest of the site works. This is confirmed to be a problem with:

- Netscape 4.05, Win95
- Netscape 4.05, WinNT
- Netscape 4.05, OSF/1 (Digital UNIX)
- Netscape Navigator 4.05 (Linux)

> Does anyone have any ideas/fixes, or is it a server-specific problem?

Browser-specific, I think; follow me on this. In testing a sample Web
page, I created combinations of URLs that matched or were similar to the
banner advertising URL used by the AltaVista mirror. The browser was
Navigator 4.05 for Linux. Results (here "NL" is newline or line break
and "CR" is carriage return .. Unix uses NL alone, DOS/Windows uses a CR
followed by NL for its end-of-line terminator):

- A/HREF OK
- A/HREF with CR OK
- A/HREF with NL OK
- A/HREF with CR+NL OK
- ILAYER/SRC OK
- ILAYER/SRC with CR OK
- ILAYER/SRC with NL OK
- ILAYER/SRC with CR+NL FAIL

For some reason when parsing the SRC= URL in ILAYER tags, Netscape fails
to strip out both the CR and NL (but either alone is fine), even though it
does strip them out of HREF= in A (anchor) tags.

The quick solution is for Telstra to "fix" their HTML by removing the line
break in the URL completely (ie., remove both CR and NL); longer-term,
though we should try and get this parsing bug in Netscape fixed.

Does anyone have Mozilla 5.0 handy to test if this still exists in the
open source code?

Cheers..

David J. N. Begley <d.begley@nepean.uws.edu.au>
Network Analyst, Communications Unit
University of Western Sydney, Nepean
Australia
Received on Fri May 29 1998 - 21:56:08 MDT

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:40:31 MST