Solved! Telling a browser to keep a connection alive

[Madison Kelly]

Hi all,

   I got a lot of feedback on this problem when I posted a little while 
ago so I wanted to post an update...

   I tried Fraser's Non-Parsed Headers trick and that gave me /more/ 
time but in the end, the connection still died if the server took too 
long to respond. It was a great piece of info though!! I suspect I'll be 
trying that trick elsewhere later.

   I tried a different web server (three in all, xs_httpd, apache2 and, 
the one that I've used for a while, Boa) and tried playing with their 
KeepAlive and Timeout settings but in the end, it could still die...

   What finally worked is:

   I had the main script call a secondary script that printed it's 
output to a file. The main program called this file using '& |' to 
disconnect and return. To be sure I didn't have a run-away script, I 
would watch for a specific temporary file to appear. If it took more 
than ten seconds than I knew the secondary script failed and I die. If 
it did show up though then I knew it launched okay and and I read it's 
contents  and printed it to the screen.

   When the secondary script starts it adds and html REFRESH to the top 
of the page. This refresh calls the main script every period of time 
passing an extra cgi variable to tell it that it's just refreshing the 
screen. Before doing this though it checks the 'mtime' of the temp file 
to make sure it's still being updated. If the file hasn't been updated 
recently (X seconds) it considers the secondary script to have died, 
kills the temp file and prints an error.

   So long as the temp file is still being updated though the main 
script will read the temp file and print it's contents to the screen. 
The last thing the secondary file does when it is finished is to re-read 
the temp file, delete the REFRESH line and re-save the file. This way 
the last time the main script loads the contents of the temp file it 
stops refreshing the screen.

   The secondary script will, after waiting 2 seconds longer then the 
REFRESH time, delete the temp file to clean up.

   Voilà! A connection between the client and the server that will not 
die. More over, the user can even close his or her browser on a big job 
and periodically reload the page to see how the job is doing (ie: a 
backup job that could take hours on very large systems).

   I am sure it could be done more elegantly but for me, it works. :p

Madison

PS - Thanks to Marc (and Cameron here at home) for suggesting the "pull 
from client method".
PPS - Thanks to Tom for the article on users and wait times.
PPPS - Thanks to everyone who replied!!

-- 
-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
           Madison Kelly (Digimer)
    TLE-BU; The Linux Experience, Back Up
Main Project Page:  http://tle-bu.org
Community Forum:    http://forum.tle-bu.org
-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
--
The Toronto Linux Users Group.      Meetings: http://tlug.ss.org
TLUG requests: Linux topics, No HTML, wrap text below 80 columns
How to UNSUBSCRIBE: http://tlug.ss.org/subscribe.shtml