From: Kirk Haines <khaines@...>
Date: 2003-10-23T05:01:39+09:00
Subject: Webrick for IOWA?

Could someone with some experience with webrick maybe shed some light on
something that I am wondering?

IOWA's basic architecture has the IOWA application running as a standalone
process.  When the webserver receives a request that is to be handled by
the IOWA application, a connection is made to the IOWA process via either
a CGI program or (currently) a mod_ruby handler.  The request is
encapsulated in a serialized object that looks and acts a lot like an
Apache::Request.  The IOWA app does its thing, and returns back to its
caller (the cgi program or mod_ruby handler) the Apache::Request with http
headers and content, ready to push on back out to the browser.

This works pretty well.  It allows the IOWA app full control over the HTTP
headers that are going out to the browser, and full access to the headers
that came in, and does all of that with an acceptable level of performance
(on my server hardware the typical apps/pages can be rendered at around
20/second).

Since webrick is now a standard component of Ruby, what I am wondering,
really, is exactly what webrick provides?  Would it be
worthwhile/practical for me to look at changing IOWA so that webrick
provides the framework for the communications conduit?


Thanks for your thoughts,


Kirk Haines