Mailing-List: contact tomcat-dev-help@jakarta.apache.org; run by ezmlm
Precedence: bulk
Reply-To: tomcat-dev@jakarta.apache.org
Date: Wed, 13 Jun 2001 11:52:31 -0700
From: Justin Erenkrantz <jerenkrantz@ebuilt.com>
To: Salim Virani <salim.virani@usa.net>
Cc: tomcat-dev@jakarta.apache.org
Subject: Re: ICAP servlet classes (was: Tomcat Architecture)
Message-ID: <20010613115231.S333@ebuilt.com>
References: <20010613113047.Q333@ebuilt.com>
 <NFBBIHLDFMOMLANBAAJFMEDCCCAA.salim.virani@usa.net>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
User-Agent: Mutt/1.2.5i
In-Reply-To: <NFBBIHLDFMOMLANBAAJFMEDCCCAA.salim.virani@usa.net>;
 from salim.virani@usa.net on Wed, Jun 13, 2001 at 02:34:17PM -0400

On Wed, Jun 13, 2001 at 02:34:17PM -0400, Salim Virani wrote:
> Hi Justin,
> 
> I faced a similar problem. I figured I could write my own server/servlet API
> .. which is what I did.
> 
> Just curious .... was your implementation in Java ??? And are you working
> for a company or this is some school project ??

We had a client who requested the use of the NetApp NetCache proxies
with ICAP.  We used Apache 1.3 as the base.  We didn't believe that 
Java was scalable enough for our needs (commercial site).  We ended 
up tossing the guts of Apache's protocol engine and came up with one 
that worked only with ICAP.  What Craig just suggested for Tomcat might
work better than what we did with Apache, but we needed high-performance 
- hence C code.  HotSpot only does so much.  =)

Roy Fielding (here at eBuilt) and Don Gillies (at NetApp) had some
interesting exchanges over how to interpret the HTTP spec.  Don 
Gillies' implementation of ICAP in the NetApp boxes was quite buggy 
and awful.  The WG's reference implementation of ICAP is a perl 
script.  And, that script isn't even remotely RFC2616 compliant.
The WG has now dropped all pretenses of making ICAP HTTP-compliant.
The next revision of the ID will have this noted.

If we had to do it all over again, we would never use ICAP.  The idea of
ICAP in and of itself isn't bad, but their implementation has no clue
what it wants to be.

I'd recommend subscribing to the ICAP mailing list
(ietf-openproxy@imc.org) if you haven't already.  -- justin