> On Monday 11 November 2002 01:51 am, you wrote:
> > Something interesting, I spotted the cookie debug options in mon.cgi and
> > uncommented them. The strings that appear at the top of the webpage are:
> > cookie value is                (never changes)
> > encrypt password is    (never changes)
> > username is ""
> > decrypt password is ""
> >
> > The last two have the correct values in them when I log in, but *only*
> > for the page following me logging in. Some more fiddling shows that no
> > matter what user/pass I type in, it accepts it for the first page, but
> > then forgets it....WTF??
> 
> 
> Since cookie value, username and password are blank, as read by mon.cgi from 
> the cookie contents, TF is, you have either a problem with:
> 1) your browser not accepting the cookie properly
> 2) your browser expiring the cookie early
> 3) your browser somehow not presenting the cookie back to mon.cgi
> 4) mon.cgi somehow not being able to parse the cookie
> 
> I would doubt #4 is the case, since mon.cgi uses very standard cookie parsing
>  
> as provided by CGI.pm. That leaves 1,2, and 3, all of which could have  
> several possible causes.
> 
> So, try a couple of different browsers on a couple of different machines. If 
>  
> they all failed, I'd be very surprised. Every piece of evidence so far points
>  
> to a client-side problem.

Even though the problem has magically disappeared I'll throw in this
possibility: might some confusion about the hostname be the cause? E.g.,
The cookie is set for host "localhost" but all the links the mon.cgi
generates may have a different hostname. Without "using the source luke"
I'd guess mon.cgi doesn't even use the hostname in links, but this gives
the flavor of the problem. Something as simple as not using a FQDN can
arse cookies up.

What I would have done is run ethereal to capture the tcp stream of a
login session from the server to the client, verifying the Set-Cookie
and Host headers are correct and then capture the stream from client to
server when going to another page, verifying that the client regurgitates
the cookie back to the server.

-- 
Mark Wagner [EMAIL PROTECTED] 206-598-0302
Unix System Administrator, Radiation Oncology and Radiology
_______________________________________________
mon mailing list
[EMAIL PROTECTED]
http://linux.kernel.org/mailman/listinfo/mon

Reply via email to