[9748] in www-talk@info.cern.ch
Re: Server API's vs. CGI
daemon@ATHENA.MIT.EDU (Paul Everitt)
Wed Jan 31 08:14:42 1996
Date: Wed, 31 Jan 1996 08:11:58 -0500
From: Paul Everitt <paul@digicool.com>
To: dperraudin@arcadis.be
Cc: www-talk@w3.org, dclc@digicool.com
Resent-From: www-talk@w3.org
dperraudin@arcadis.be wrote:
>
> I have been following the discussions regarding the various server API's that
> are avaialble. What kind of performance improvement could one expect from
> using server API's rather than CGI scritps?
Using our API scripting stuff:
http://www.digicool.com/releases/
the performance bump looks quite good. _Preliminary_ testing showed a 250%
drop in latency for similar operations. Throughput should be even better, as
the machine doesn't have new processes to deal with. Please note that there
were a lot of caveats in the testing.
We are investigating some other techniques to give a significant bump, past
reading an HTML file from disk.
> Is any group attempting to define
> a standard set of server API's? Is there a FAQ available on this topic,
> or a website with more information.
I am trying to get caught up on all the current proposals, do some research,
and field a draft. Our assertion is to have a mechanism that is:
a) httpd-independent: differences between APIs should be, at a base level,
hidden, but available for extensions
b) platform-independent: obviously, OS shouldn't matter
c) language-independent: big point here. If it requires C, as current APIs
do, then a vast audience is lost. A language-neutral stance, as is CGI,
would be best.
d) location-independent: partitioning operations across machines should be
allowed, if possible
e) use of industry-standards and well-known technology
--Paul
