The Fortran-filter Gateway

[Mark Boolootian <[e-mail]>]

[Ran across this on the RISKS Forum and thought you might find it amusing - mb]


Date: Fri, 26 Mar 93 23:04:46 HST
From: "Joe Dellinger" <[e-mail]>
Subject: The FORTRAN-hating gateway

	Several months ago we started noticing that (now and again) the
network connection to the mainland would become very very slow; this would
continue for 10-15 minutes or so, then all would suddenly be well again.  A
while after this started happening a coworker of mine complained to me that
the connection to the mainland _never_ worked anymore. It seems that he had
some FORTRAN source that he needed to copy to a machine on the mainland, but
he never could because "the network wouldn't stay up long enough for the ftp
to complete".

	Yes, it turned out that the network outages happened whenever he
attempted to ftp that _particular_ FORTRAN source file to the mainland. We
next tried compressing the file; it copied just fine then (but unfortunately
the machine on the mainland had no uncompress program, so it was still no go).
Finally we "split" his FORTRAN program up into very small pieces and sent them
one at a time. Most of the pieces would copy without trouble, but a few would
either not go at all or only go after many _many_ retries.

	Examining the troublesome pieces, we found they all had one thing in
common: they contained comment blocks that began and ended with lines
consisting of nothing but capital C's (his preferred FORTRAN commenting
style). At this point we started sending e-mail to the network gurus on the
mainland asking for help. Of course, they wanted to see an example of our
un-ftp-able files, so we mailed some to them... but our mail never got there.
Finally we got the bright idea of simply _describing_ what the unsendable
files were like. That worked. :-) [Dare I include in this message an example
of one of the offending FORTRAN comment blocks? Probably better not!]

	Eventually we were able to piece together the story. A new gateway had
recently been installed between our part of campus and the connection to the
mainland. This gateway had GREAT difficulty transmitting packets that
contained repeated blocks of capital C's!!!! Just a few such packets would
occupy all its energies and prevent most everything else from getting through.
At this point we complained to the gateway manufacturer... and were told "Oh,
yes, you've hit the repeated C's bug! We know about that already.".
Eventually we solved the problem... by buying new gateways from another
manufacturer. (In the manufacturer's defense I suppose an inability to
propagate FORTRAN programs might be considered a feature by some!)