+- The Un-Official Proxomitron Forum (https://www.prxbx.com/forums)
+-- Forum: Forum Related (/forumdisplay.php?fid=37)
+--- Forum: Archive (/forumdisplay.php?fid=50)
+---- Forum: Proximodo (/forumdisplay.php?fid=40)
+----- Forum: Issues with Proximodo (/forumdisplay.php?fid=27)
+----- Thread: Outbound Connections (/showthread.php?tid=335)
- Guest - Dec. 05, 2004 03:36 PM
I was wondering how Proximodo handles HTTP/1.1 persistent connections?
If my browser has 8 persistent connections per server set up it will make up to 8 connections to Proximodo (irespective of what is loading as to the browser the proxy is the only server it knows right?).
BUT what does proximodo them do - is it purely transparent or does it then only make two persistent connection per external server (as HTTP/1.1 recommends)?
Also, how does it handle HTTP/1.1 pipelining - is this again transparent to the proxy?
- Guest - Dec. 05, 2004 05:48 PM
A little update:
Using a benchmark page with 12 inline elements, Opera V7.6PR4 opens 4 persistent TCP connections without any proxy (checked with Ethereal). With proximodo, it opens 4/5 connections, and with Proxomitron (V4.5j) it always opens 8 connections (some carrying very few bytes in them).
So it seems (at least on a test page) tht Proximodo is more transparent/efficient than Proxomitron is!!! B)
When I have time, I'll test off a local server, and see which delays the HTML packet more during processing. Has anyone thought of making a test-suite for this kind of testing?
- kuruden - Dec. 05, 2004 08:11 PM
Currently, 1 website connection for 1 browser connection. Http recommendations are good, and will be implemented some day, but it's a lot of (re)work. Wanna do it?
Doing a test suite is a great idea.
- Guest - Dec. 05, 2004 11:52 PM
I don't code
But I think Proximodo's current behaviour is pretty good. It should not impose restrictions by default - if the browser uses 4 connections, I suppose proximodo should oblige? I normally use a standards page to test browser loading efficiency. It is a longish doc with 1 .js and 1 css + 10 images, 1 called via CSS. It can sit happily on a local server. It has two javascript timers built-in: http://nontroppo.org/test/timer/
The problem with a test suite is that for header filters a HTML test suite is meaningless - PHP could be used to manipulate sent headers in those cases...