httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Brian Behlendorf <br...@hyperreal.com>
Subject Re: WWW Form Bug Report: "PassEnv doesn't pass some variables" on NeXT (fwd)
Date Wed, 18 Sep 1996 17:49:30 GMT


---------- Forwarded message ----------
Date: Wed, 18 Sep 96 10:39:24 -0500
From: Bill Bumgarner <bbum@friday.com>
To: Andrew Wilson <andrew@aaaaaaaa.demon.co.uk>, robh@imdb.com
Cc: new-httpd@hyperreal.com
Subject: Re: WWW Form Bug Report: "PassEnv doesn't pass some variables" on NeXT


Nope-- not a bug; I'm an idiot (I failed to remember that three of the four variables I did
a PassEnv on were already passed in).

BUT

It would be extremely valuable to be able to pass various variable values found in the configuration
files into the cgi-bin's environment in an automated manner.  IE;  if I say 'PassEnv ServerRoot',
it should pass the ServerRoot configuration value into the cgi-bin's environment.

Along the same lines,  I find working with the configuration files to be very frustrating.
 There are two major 'features' that are completely missing or horribly confusing.  Because
of these anti-features, it is extremely difficult to put together a set of customizable, modular
configuration files that allow one to isolate all of the system/site dependent components
such that moving the site (between systems or within the parent site) is not prone to introduce
lots of errors.

First:  There is no concept of assignable variables within the configuration files.  Sure;
 one can set various variables that control the operating characteristics, but one cannot
set a variable whose value can be used throughout the configuration process.  

For example, once I have defined ServerRoot-- either for the master server OR for any given
VirtualHost-- I would like to use the ServerRoot value to construct other paths.   Some paths
work this way-- others don't.  For example, the log file directives work great.  BUT  <Directory
...></Directory> blocks do not.

Instead of having some paths automatically pre-pend the ServerRoot value and others not, it
would be preferable to have some simple syntax for accessing variable values-- say, ${ServerRoot}
or something equally as shell/perl (blech!) like.

Not only would this greatly increase the flexibility of the configuration files, it would
also greatly improve the site administrator's ability to produce and maintain a quality site
without having to constantly twiddle paths throughout the entire site.

---

Secondly;  I have found the relationship of variable values between configuration files and
within VirtualHosts to be confusing, at best.  Specifically;  there is no documentation as
to how the variable space of a virtualhost is derived from the variable space of the master
server's configuration.

To aid in modularity, I have defined my virtual hosts such that they have an access.conf and
srm.conf of their own-- It appears that they inherit some, but not all, of the master's configuration...


Ideally, I would like to be able to add a single, simple VirtualHost entry in httpd.conf--
with custom access.conf and srm.conf files within the VirtualHost's directory and have everything
+just work+.  I have come close, but the need for hard-wired paths all over the place is killing
me-- what if I don't want every single web server on my WAN to be in /usr/local/httpd (I hate
/usr/local/etc/httpd)?  Maybe my production staff needs a web server tucked away in a different,
internal part of the file system-- yet I want to be able to basically move a tree of resources
from the content development server into production without having to modify a bunch of hard
paths!

b.bum


Mime
View raw message