On Friday, August 18, 2006 05:05:46 PM -0500 Nicolas Williams
<Nicolas.Williams%sun.com@localhost> wrote:
I suspect that the set of people likely to be interested in working
on a
filesystem access protocol is different than the set where were
interested in working on the core ssh protocol. So, it's entirely
possible that a BOF on this subject may produce enough interest to
form
a WG.
For one item? Note that the security area wouldn't be a very good home
for this. The transport area seems more appropriate, particularly for a
filesystem protocol.
It's only one item if you make it huge and monolithic. If, instead,
you break it out into a manageable base document and a set of optional
extensions, it becomes several items. It also offers the opportunity
to take a step back and figure out what the goals and requirements
actually are. So far, it seems like they get rehashed with every
proposed extension, and it's starting to get hard to tell what the
consensus is. About the only thing that's clear to me is that the
people working on this project mostly are interested in something like
a filesystem, rather than something like a file transfer protocol.
I agree the security area is not the right home for this. I'm
inclined to think the right home is actually the apps area; I've never
understood why NFS was in transport and not apps. Perhaps someone
else knows.
-- Jeff
Some have argued that the recent additions to the SFTP draft make it more
like a filesystem.
I would argue the basic implementation that most folks are using (SFTP v3)
is more like a filesystem than a file transfer protocol. In particular,
there is no put or get in SFTP v3 - instead it has open, close, read,
write, etc.
I'm not sure where the work or WG belongs, but I would like to see the SFTP
draft completed (acknowledging that is has always been more filesystem than
file transfer) - even if we had to strip it down to core functionality and
push off some of the newer stuff to extensions.
I'd also be very supportive of a file transfer draft too. However, I'm not
sure there is enough consensus to make this happen :-) I think having a
documented standard for transferring files using more of a put and get
mechanism would address many of the real world issues (e.g. performance and
logging) that SFTP has.
--Jeff V.