Is HTTP necessary??

marca@ncsa.uiuc.edu (Marc Andreessen)
Date: Mon, 17 May 93 18:45:55 -0500
From: marca@ncsa.uiuc.edu (Marc Andreessen)
Message-id: <9305172345.AA01779@wintermute.ncsa.uiuc.edu>
To: p.lister@cranfield.ac.uk
Cc: ccprl@xdm001.ccc.cranfield.ac.uk, www-talk@nxoc01.cern.ch
Subject: Is HTTP necessary??
In-reply-to: <9305171402.AA14823@xdm039>
References: <9305171402.AA14823@xdm039>
X-Md4-Signature: ab967f10b900e2de6e5d56e45d3bfd85
"Peter Lister, Cranfield Computer Centre" writes:
> The mapping functionality of httpd can be useful, but does it do
> much that can't be achieved by bunging symbolic links in a Gopher
> directory?  OK, so one can map different filename formats and that
> has been useful here, if not vital. However, directory browsing is
> *not* nice with httpd if I have to tag every directory with
> .www_browsable; I want inexpert users to be able to publish
> information, and I forsee problems if they need to do anything
> "unexpected" like that.

That is one particular httpd -- at least one other that I know of
(NCSA httpd) and I assume others (BSDI's, etc.) always support
directory browsing.

> HTML is great, but is HTTP worth it? I'm want practical details, so
> don't send me philosophical waffle or flame me for being a traitor
> to WWW, cos I'm not. 

HTTP is very lightweight and very simple, which facilitates extremely
fast server implementations for document query and retrieval.  This
will be significant for serving many people simultaneously vs Gopher
or FTP.

It can also be argued that HTTP is more flexible, as servers can be
trivially constructed to handle a wide variety of specialized data
retrieval functions.  (For instance, a colleague wrote an HTTP 'xgraph
server' the other night that accepts URL's that contain multiple sets
of {x,y} coordinates and returns documents containing inlined images
that have the points plotted out using the guts from xgraph for the
graphics -- try that with the Gopher protocol.)

> Yes, I know that there are rumours that Gopher is to become
> proprietary. Unfortunately, while I regard that as bad, others here
> may regard it as *good*; the idea of paying for a support contract
> makes them feel warm, and secure, despite what experience should by
> now have taught us.

Well, this is a whole 'nuther issue.  The only thing I want to point
out right now is that there's no reason you can't pay for a support
contract for any protocol/server, including FTP, NNTP, or whatever.
And besides, it shouldn't have to be an either/or proposition -- look
at freeWAIS (CNIDR) and WAIS Inc.  And besides, Gopher will likely
still be freely available to non-commercial users, and besides, if
not, others will be doing Gopher servers that will be (if Gopher is in
fact worth doing this for; no comment from me on that one).  And
besides, blah blah blah...

Cheers,
Marc

--
Marc Andreessen
Software Development Group
National Center for Supercomputing Applications
marca@ncsa.uiuc.edu