At the risk of beating a dead horse; a few links to related (and
fixed) bugs in other projects (including firefox).
The root cause of the hack (Dan just touched on it):
https://bugzilla.mozilla.org/show_bug.cgi?id=246651
Message reporting that the above bug was fixed in the 11 Sep 2004 build :
http://forums.mozillazine.org/viewtopic.php?p=788974
Two bugs from prototype (the second is marked duplicate):
http://dev.rubyonrails.org/ticket/5606
http://dev.rubyonrails.org/ticket/5809
With this info, the discussion we have already had in this thread,
Jake's beforeSend tip and a bit of luck, we might be able to get away
with one more message on the subject. One saying that the "close"
code/work around has gone away.
Ian
On 4/18/07, Dan G. Switzer, II <[EMAIL PROTECTED]> wrote:
Jörn,
>if ( xml.overrideMimeType )
> xml.setRequestHeader("Connection", "close");
>
>I'm still not sure when that is really necessary to set. Could someone
>summarize the conclusion from this thread?
In older Gecko builds v1.0.x, the content length was incorrect unless you
passed in the connection close instruction. This meant that your AJAX
packets would get corrupted.
-Dan
