Hi Ray -
Thanks for the response. See a few comments below.
At 05:56 AM 4/7/2015, Ray Pelletier wrote:
Michael,
On Apr 6, 2015, at 12:14 PM, Michael StJohns
<mstjohns(_at_)comcast(_dot_)net> wrote:
At 01:31 PM 4/3/2015, IETF Administrative Director wrote:
All;
The use of FTP to retrieve files from the IETF servers has been
declining steadily.
Hi Ray -
What is the cost of leaving FTP service in place? I would expect that the
cost was more in setting up FTP in the first place (e.g. adding it to
template emails, configuring the servers) than any real day to day cost, but
I could be wrong.
What is the cost of removing it? (e.g. besides simply shutting off the
servers, what else has to be done, who has to do it and what will it cost?).
Is there an impact on the IETF budget either way short term? Long term?
Thanks - Mike
There is no cost to turning it off.
As someone else noted, there are a number of ftp:// links scattered throughout
the IETF database. There will be a cost to go through and re-write the
templates or even re-do documents. Probably not a great amount, but those
costs will exist.
There is an incremental savings in maintenance and support burden,
and complexity.
The FTP server we are using (proftpd) places restrictions on how we
store files in the file sytem that are much more constraining than
the http and rsync daemons. Essentially, the files to be served must
be stored in a single tree (as hardlinks - symlinks won't work).
This is impeding work as we evolve. In particular, it affects how we
separate services to take advantage of cloud architectures. Other
servers have different limits, but still place significant
constraints on what we can do. Since the information is already
available through other mechanisms (particularly rsync), the folks
studying the problem recommended discontinuing the service, rather
than investing in finding the least onerous deamon and reconfiguring
to it.
That makes a reasonable amount of sense. Sort of. I think elsewhere you
stated that the traffic towards FTP is decreasing and we're not talking about
adding new types of data to the set so maybe leaving this as a legacy system
with no improvements planned for a couple more years might make sense.
I don't actually have any strong preferences one way or the other - except that
I'd prefer a measured and long phase out rather than an abrupt turn off if
that's where we end up going.
Later, Mike
Ray
The files made available with that protocol are also available using
http and rsync. (See the modules exposed at rsync.ietf.org using
"rsync rsync.ietf.org::")
The majority of the current FTP traffic appears to be from mirror
sites that would be better served using rsync.
Input received by 20 April will inform the decision.
Thanks
Ray