On Wednesday 28 August 2002 22.07, Daniel Barron wrote:
> Always 1128 bytes no matter what I put in the error page.
Then the next step is to look at some packet traces..
> I agree that it's probably a bug in IE, but in the real world lots
> of people use IE and this bug happens with the very latest IE.
> Coders of web appliances such as myself have to bow to the majority
> and include work arounds for buggy browsers. Don't take me wrong -
> I mean no offense!
None taken.
> Have you any other ideas of how to solve it? Is squid possibly
> closing the connection before its flushed? Or is it that IE can't
> handle an error more than 1128 bytes on SSL?
Most likely IE is at fault here.. but I agree that there is a slight
chance that it might be a connection management issue. A packet trace
will tell.
If IE blindly assumes the CONNECT was successful and starts sending
the SSL HELLO message before receiving the CONNECT response by the
proxy then maybe there is something Squid can do to workaround IE.
If it simply is that IE does not like errors in response to CONNECT
then I am not sure..
> I suppose I could make a small error page that has a meta tag that
> redirected the browser to an error page on the web server and
> include a check for that destination in the squid.conf so it
> specifially allowed it and did not create a loop... Don't you just
> love the quality of microsoft code (!)
Only to instead get bitten by the IE "Show friendly HTTP errors"
feature... workaround one bug only to end up in another misfeature.
Note: Squid-2.6.DEVEL has the ability to send redirects on deny_info,
allowing you to try this approach. But be warned that 2.6 is not
exacly production quality.
Regards
Henrik
Received on Wed Aug 28 2002 - 15:48:55 MDT
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 17:09:52 MST