[Xitami] simultaneous frames - 50+ simultaneous users...

[Ayhan Tuncer]

Hello.


This posting refers to :

  > Subject:  Re: [Xitami] 50+ simultaneous users...
  > Date:     Thu, 14 Mar 2002 17:20:48 +1300
  > From:     Ewen McNeill <ewen@imatix.com>
  > Reply-To: xitami@lists.xitami.org
  > To:       xitami@lists.xitami.org

  in reply to:

       > Subject:   [Xitami] 50+ simultaneous users...
       > Date:      Wed, 13 Mar 2002 22:56:39
       > From:      "Thomas J. Hruska" <shinelight@shininglightpro.com>
       > Reply-To:  xitami@lists.xitami.org
       > To:        xitami@lists.xitami.org


I've had a different problem which may have a similar cause.


One of my perl cgi programs serves multiple responses against extra path
info in the url.
Responses are all html pages. All output is as expected, I have no
problems.

Except for one response which is multiple-framed:
  One frameset / 3 frames
  Javascript - no errors known to me.
  No html code errors known to me.
  No html code framesetting errors known to me.
  Html output varies - have 15 different frameset content for same
frameset scheme.
  Minimum total frameset content 11 Kb  / maximum total frameset content
73 Kb.


Problem:
--------
- The third frame locks Xitami up - regardless of content served,
- corrupts any files ( db temp lockfiles ) being written by the
  cgi program ( it executes ) during the lockup and
- crashes the computer.
Db is simple text files - no engine - no server.


Logfiles show
-------------
- Three successive requests - one frameset, two frames. Fourth request
not there.
- Request order of frames varies and is identical in requests both from
Netscape and  Explorer.
- Different content has different frame request order.
- Both browsers trigger the crash.


So that
-------
- A different page is being served during each crash,
- a different frame of the set is being served during each crash,
- the html/Js content served is different during each crash,
- the file size being requested is different during each crash,
- and a different browser is making the request at each crash.


The only thing I have found to be in common is the fact that a third
frame is being requested.

It's difficult to reproduce the environment unless I give a few hours of
time, because I changed the third frame to a Js write from the frameset
<head> content and solved the problem for good. However, I am upset not
being able to serve a third frame from a cgi program.

I use Windows 98 on a Pentium II 64 Mb Ram and Xitami 2.4d6 to serve an
introvert website. No workload - mostly testing.