Re: suggested libWWW architecture

"Tony Johnson (415) 926 2278" <>
Date: Wed, 13 Jan 1993 22:44 PDT
From: "Tony Johnson (415) 926 2278" <>
Subject: Re: suggested libWWW architecture
X-Envelope-To: www-talk@nxoc01.CERN.CH
X-Vms-To: IN%""
X-Vms-Cc: TONYJ, in%""
From Dan:

>/* I think the WWW file: should be split into ftp: and local-file:.
>   It's cleaner to implement; there are precedents in the MidasWWW local:
>   scheme and the MIME ftp and local-file access-types. */

I agree.  Here are some comments along the same lines I sent Tim a while back.

From Tim:
>Re "local:"
>The actual implementation of "file": in WWW checks the hostname to see
>whether it is the local host, and if so uses direct file access.
>(It also checks /Net/<hostname>/<path> in case of NFS mounting)
>That is why it was called "file:" rather than "ftp:".  This means that you >can
>edit files under /pub which other people get with ftp. For example.
>The IETF WG wanted "ftp:" to be available, so I have made that a synonym >for

My immediate reason for implementing local was that MidasWWW implements "file:" 
through the gateway, and so I needed some other temporary way to access local 
files. When MidasWWW supports "file:" directly I may get rid of local (or make 
it a synonym for file maybe). However I do have some other problems with the 
file implementation, mainly the confusion between files access by ftp and files 
accessed locally. Problems are:

a) How to tell whether a specified node is the "local" node? The /NET test is 
not portable, it doesn't work on our Unix system here (we use /nfs here). It 
works even less well on VAX clusters.

b) Even if I could tell that the file is on the local node, things still would 
not work since /pub is NOT the root of the anonymous FTP directory. I.e. 
file://localhost/test.html is NOT the same as local:/pub/test.html. Again this 
applies even more to non-unix systems.

Both of these problems tempt me to keep local:, or some equivalent. Comments?