ftp on triples

[Michael Barr <barr@triples.math.mcgill.ca>]

I did not go into details on the problems we had, but so many have written
to me offering diagnoses incompatible with the facts that I figure it is
easier to describe them once and for all.  There were two symptoms.  First
triples essentially stalled for minutes at a time.  More seriously, the
computer centre made a formal complaint that triples was generating so
much traffic that it was seriously compromising the internet connection.
Since McGill is the gateway to all of Quebec, this is serious business.
There was an implicit threat to cut the department off if we didn't do
something about it.  I suppose then the department could have cut triples
off, although it never went that far, since I first tried changing the
root password in case whoever it was was changing the logs.  This had no
effect and the logs left little doubt that the only oddity was the ftp
signons which showed simultaneous (maybe 8-10) from the same origin and
that not identifiable by nslookup.  (This last fact is not, in itself
suspicious; the machine I am now using gets a new IP address whenever it
is booted and cannot be identified by nslookup.)  But the machine that
does this is different every time.  So it cannot be explained by,
according to one suggestion, some newbie getting confused and copying the
same thing many times.  In fact, it is hard to know how you could copy
enough to fill the while I-net channel.  At any rate, closing the ftp
access seems to have cured it.  We are mulling over what to do.  Some
think that putting a password will inhibit non-category theorists that do
not subscribe to the list.  And I guess it would compromise Hypatia.  But
I think Vaughan mirrors triples anyway.

Michael