Date: Wed, 2 Mar 94 17:52:34 +0100
From: Tim Berners-Lee <timbl@ptpc00.cern.ch>
Message-Id: <9403021652.AA06442@ptpc00.cern.ch>
To: uri@bunyip.com
Subject: FTP URL mapping
It has been pointed out that the mapping of FTP URLs onto
FTP protocol is not well defined. In fact it seems we
can arrive at a definition which allows current methods or
working, with fall-back wallowing URLs to be created for things on
flat file systems like VM, and non-unix hierarchical systems.
What do you think of the following, folks?
__________________________________________________________
The FTP protocol allows for a sequence of CD (change directory) commands
prior to a RETR which actually accesses a file. The arguments of any CD
commands are successive segment parts of the URL, and the filename argument
to the RETR command is the final segment of the URL path.
Note
In the case in which the file system of the server is known or guessed, the
path may possibly converted into a filename. This may allow the file to be
retrieved in one command. In the case of unix, the filename will look the
same as the path. This must NOT be taken to indicate that the URL is a unix
filename. In practice, as many FTP servers in fact have or emulate unix
file systems, it may in fact be time-efficient to attempt first a direct
retrieval guessing unix syntax, and, if that fails, to attempt the official
sequence of succession of directory changes followed by a RETR command.
There is no common hierarchical model to the FTP protocol, so if a directory
change command has been given, it is impossible in general to deduce what
sequence should be given to navigate to another directory for a second
retrieval, if the paths are different. The only reliable algorithm is to
disconnect and reestablish the control connection. However, if no directory
changes have been made, but direct retrieval has been done, then the control
(This note previously read: "The adoption of a unix-style syntax involves
the conversion into non-unix local forms by either the client or server. Some
non-unix servers do this, but clients wishing to access sites which do not
have unix-style naming will need certain algorithms to enable other file
systems to be identified and treated. Client software may also have to be
flexible in terms of the sequence of FTP commands used with different
varieties of server. In view of a tendency for file systems to look
increasingly similar, it was felt that the URL convention should not be
weighed down by extra mechanisms for identifying these cases." )
Tim Berners-Lee