Re: [Trinityos-perlwebmin] What's up ?
Status: Alpha
Brought to you by:
ep98
From: <tom...@us...> - 2002-06-25 23:13:05
|
Stefan Ilivanov wrote: > It's there any progres with the administration system ? I've brought forward the I/O utilities code which is part of the webmin base about 18 sub releases. It includes an implementation of C flock(3) in pure Perl which isn't portably available and is error-prone in the core of the Perl language itself. I've been planning something out which I probably should present for public approval before going forward with it anymore. I already have figured out how to implement this proposed action, so this isn't a pie-in-the-sky idea. It's a sub-sub project looking for approval and feedback. I propose that a seperate apache server be created to specifically handle the web admin. It should listen on a different port, and be compiled differently than the main thttpd web server. There are several reasons this should happen, and several reasons why it is beneficial. I'll go over these from this point forward. Due to the nature of a dedicated server, we can profit from tactics which wouldn't work for the main server. (read on) The seperate apache binary can be a mod_perl server and still be a superfast, streamlined process. Because only handles a few documents and CGIs it keeps its registry optimally small. There's nothing faster than RAM, folks. The dedicated server can be compiled with the inclusion of the base webmin code and its supporting libs without losing performance; even GAINing performance by a significant measure. The dedicated server can be compiled with the EXclusion of more apache mods which would otherwise be neccessary on a fully-functionally public apache server. (eg- we don't need to compile it with a bunch of language packs which aren't used as the OS native language, no need for logging when base code already handles this much better with a searchable interface, no need for huge mime db, no need for server-side scripting support other than the perl, AND a whole bunch of other stuff.) The seperate httpd is going to be very transparent and have a verrrrry small memory footprint while demanding fewer cpu cycles. The dedicated server can be safely limited to a small number of child processes. A seperate apache binary allows us to avoid getting muddled by, and muddling up further the public web server with all its issues. We remain clear of the main server's operations that have nothing to do with us: load balancing policies for public documents shouldn't interfere with us Think about it for a moment: how can web admin fix a hanging server if it RUNS ON IT? We don't need the same storage schedule, we don't need to run the same cron jobs, and the pub server really doesn't need to be messing around with ours! --especially if the inroad I'm working on provides a further layer of security between the web admin and the web browser by putting risky jobs in que for later admin approval with strictly onsite system access. I want my own crontab. Any special firewall rules for our server don't need to slow down the main server, and vise versa. Our logging semantics and subprocesses are different. We don't need to run extra helper procs and subprocs for our own programs. We don't need to include our access hits in the traffic reports, for a simple example. We need different levels of access to the machine's internals which we shouldn't have to build into the main server. We can easily target and eliminate intrusion attempts on the web admin (yes, they will happen) providing more security by staying away from the main server. If at any time we need or decide to use SSL, we're going to have to create another server anyway. This is gonna be a super pain in the butt if we have already built the webmin to work as part of the public server. This is yet another conflict we don't need. -- -Tommy Butler http://atrixnet.com I'm looking for work! contract or perm, full or part time Download my résumé http://www.atrixnet.com/resume the ooOPps Code Library http://ooopps.sourceforge.net/pub/ |