about WebNFS

Rohit Khare (khare@w3.org)
Wed, 29 May 1996 12:58:01 -0400 (EDT)

A quick analysis from our Java guru....

Resent-Date: Wed, 29 May 1996 12:54:29 -0400
Resent-Message-Id: <199605291654.MAA23454@www19.w3.org>
Date: Wed, 29 May 1996 12:53:52 +0500
From: abaird@w3.org (Anselm Baird-Smith)
To: w3t@w3.org
Subject: about WebNFS
X-List-Url: http://www.w3.org/pub/WWW/Archives/Team/w3t/
X-See-Also: http://www.w3.org/team/WWW/Team/
Resent-From: w3t@w3.org
X-Mailing-List: <w3t@w3.org> archive/latest/1559
X-Loop: w3t@w3.org
Resent-Sender: w3t-request@w3.org

I just went through the WebNFS white paper:
[don't print it reuse my own copy]

My understanding is that they just made some improvement in the
protocols (MOUNT and NFS per-se) to make it run smoothly on WAN
(it mainly eliminates some round-trips).

However, the expected drawbacks are there too: NFS is just a file
transfer protocol, while it's my understanding that most of HTTP
cleverness lies in its extensible set of methods. Nothing is said
about how CGI scripts (eg POST) or content negotiation, etc. would be
handled for example. Meta-information is of course out of the picture
(e.g. to get content type, parse the file suffix).

I hope that we will get HTTP-NG (or at least MUX) out fast enough to
avoid these kinds of things, whose main argument is that HTTP as
currently specified is just not fast enough.