Mailing-List: contact dev-help@httpd.apache.org; run by ezmlm
Precedence: bulk
Reply-To: dev@httpd.apache.org
Date: Mon, 22 Oct 2001 14:19:27 -0700
From: Daniel Lopez <daniel@rawbyte.com>
To: dev@httpd.apache.org
Subject: Re: New proxy hook
Message-ID: <20011022141927.A2295@rawbyte.com>
References: <20011022132844.A20346@rawbyte.com>
 <B7FA3F72.EE67%pier@betaversion.org>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
In-Reply-To: <B7FA3F72.EE67%pier@betaversion.org>;
 from pier@betaversion.org on Mon, Oct 22, 2001 at 09:26:10PM +0100


> > I understand your concern with easy configuration. What I am advocating is
> > adding all those features but still use HTTP as the transport (like
> > subversion does for cvs). This eliminates the need to develop and maintain
> > your own protocol, allows you to take advantage of things like SSL and any
> > improvements that you make to reverse proxy, like load balancing, can be
> > used by other modules or setups with any backend server, not only tomcat.
> 
> And you keep missing my point when I say that you can already do that with
> Tomcat 4.x, and WARP/WebApp are an addition to that method in terms of
> performance and ease of use... We already have and use the functionality
> provided by an HTTP-based reverse proxy (TC4.0/4.1, Tomcat HTTP/1.1
> connector and mod_proxy in Apache), but we need more :)

I want to understand your point :) 
What I understand you are saying is:
a) Yes, you can do that with mod_proxy, but we need more so...
b) ... we create warp connector to deal with those things that we need

What I say is, instead of creating a replacement for mod_proxy, build those
things that you need on top of mod_proxy (like load balancing, ability to
read configuration remotely, etc.)

In any case, that is just a suggestion, based on my experience. The beauty
of software is that there are many ways of doing the same thing :)


The point of my original email was to add a hook to mod_proxy to allow
headers to be modified when talking to a backend server. I got a +1 from
Chuck and some concerns from Graham suggesting if that could be done
modifying headers_in. I think I have addressed those concerns and would like
to ask for others +1 or additional concerns.

Thanks

Daniel