Date: Sat, 5 Mar 94 09:21:31 CST
Message-Id: <9403051521.AA23619@boombox.micro.umn.edu>
From: "Mark P. McCahill" <mpm@boombox.micro.umn.edu>
To: uri@bunyip.com
Subject: another way of thinking about ftp URLs
Looking back at the last six months of discussion of the ftp URL,
it seems like people are coming at the problem with significantly different
functional requirements or design goals for the URL.
Some people would like an ftp URL that is easy to create (author) and here
little burden is put on the author of the ftp URL to specify exactly
what is being pointed at or how to resolve the location. This is nice for
authors of URLs since they have less work to do, and not so nice for ftp
URL consumers since they have to have clients that are a lot more clever
about how to resolve the location... or if they don't have a really clever
client the URL consumers are going to have a personal involvement in giving
hints to the client software about what to try next.
Other people would like an ftp URL that puts more burden on the creator of the
URL to carefully specify how to actually resolve the reference to this specific
item on this specific ftp server. This is not so nice for the authors of the
ftp URLs since they have to do the work of seeing what sequence of commands
will work with this specific server. Presumably, the URL's author has already
done this work to actually get to the object being referencing, but still, this
is more work. If the authors of the ftp URLs pass the knowledge they have of
specifically how to resolve the referece on to the consumers then the consumers
have an easier time of it.
Either way, someone has to understand the details of how the specific ftp
server works since the recent discussion has made it clear that there is a
lot of variance in implimentation of server. So... the question is "Who
does the hard work?"
If you think there are going to be more URL consumers than authors, you
probably lean toward making the authors do more work. You also hope for a
syntax that has room to say in detail which sequence of commands will work
to for this specific location. If you think that authoring should be easy,
then you put the burden on the consumers of the ftp URL, and client software
for the ftp URL needs to know a lot about different implimentations of ftp
servers.
I think there will be more consumers of ftp URLs than authors, so I hope
for some sort of syntax that makes the author specify in detail what
sequence of commands should be issued to get to the item. If that doesn't
happen, then the clients are going to be a lot more complicated to develop.
Anyway, the real question is
"Should the producer of the ftp URL pass on the information that they
presumably already have (from having actually fetched the item)
or should the consumer of the ftp URL be required to re-create this
knowledge?"
Once that design decision is made, its easy to come up with a syntax that
accomodate the design goal.
Mark P. McCahill
gopherspace engineer/University of Minnesota
mpm@boombox.micro.umn.edu
612 625 1300 (voice) 612 625 6817 (fax)