Subject: | |
From: | |
Reply To: | |
Date: | Tue, 25 Feb 1997 07:48:22 -0800 |
Content-Type: | text/plain |
Parts/Attachments: |
|
|
Regarding the ongoing browser hangs, Neil writes:
>Thanks, Ken.
>We suspected incomplete html, but a thorough check and small test page
>proved otherwise - certainly incomplete html may cause this problem. We
>will check the bytestream thing as well, but in the meantime:-
>
>We have also found that bypassing our proxy server for the Intranet URL
>(Apache on the HP3000) seems to cure this problem.
>
>On MS IE 3.01, it can be done by View/Options/Settings/ then check the
>box do not use proxy server for local addresses, and enter the address
>in the box.
I wanted to mention some things I discovered when writing QWEBS which
might help the APACHE/etc. porting efforts regarding this issue. If they
apply then good, if not then feel free to ignore.
1. When using a proxy server the entire http message would (most of
the time) never get read by the server's socket read. Another socket
read would be required to fully capture the complete data sent
by the browser client.
I traced this down to the extra headers that get appended when
the message is forwarded by a proxy server. The http message
gets large enough to not get read entirely by the first socket read.
This also caused large form transmissions (usually with POST) to
require multiple socket reads to get the entire message.
2. Changing a socket from wait to nowait midstream would offend some
browsers and not others. The most common result would be a connection
reset detected by the browser.
Good luck!
Duane Percox ([log in to unmask] v/415.306.1608 f/415.365.2706)
http://www.qss.com/ http://www.qss.com/qwebs
http://www.qss.com/faq3k http://www.qss.com/qsdk
|
|
|