Jim, please forward this to the CREN board. Note the 'Reply-To:' field.
This is my last attempt at being civil, if not diplomatic, in my
description of this preposterous self-imposed restriction that I am tired
of seeing "added to the list of problems deserving future consideration"
or otherwise ignored. I am not interested in being told once more about
the details of the internal CREN bureaucracy that have to be followed in
order to open up a possibility for an eventual change of this limit. It
simply has to be removed - not "some day", but now.
LISTSERV is being rewritten in a compiled language. As a result, a large
number of small REXX files are slowly being replaced with a larger
executable file. As the REXX code is converted, the size of this file
increases, and it has now reached the Holy Limit of 300kb (which really
doesn't mean that much code). The result is that some people, connected
behind sites that enforce the 300k limit in an attempt to be "good
citizens" (and how can they be blamed for following official,
rubber-stamped recommendations?) are no longer able to receive fixes from
me, and indeed they will not be able to install new versions of LISTSERV
until this limit is removed or I send them the necessary files via other
means (FTP). I am not going to change all my carefully tuned and tested
distribution procedures just for a ridiculous historic limit which CREN
should actually be ashamed of not having addressed yet. I am doing enough
volunteer work to help BITNET stay alive to resent being given answers
which boil down to "Yes, we agree it is a problem but we have decided
that it is urgent not to do anything about it at the moment lest we might
do something we come to regret later".
Some of you may remember that I was the first to advocate enforcing file
size limits - in 1987, when all the traffic between Europe and the US had
to fight for access to two half-duplex and unreliable 9.6k links. I
actually wrote a popular package to hold large files during prime time
and release them at night, and a series of traffic recommendations to
improve throughput and user service (known under the codename E87TRAFF).
But these recommendations already said that the maximum file sizes should
be increased as faster lines become available. In 1987, I was
recommending a limit of 10k records, or 780kb. We are in 1992 now, and
there is a 768k line just for traffic between the US and Scandinavia,
while CERN enjoys twice as much bandwidth. The BITNET II core in the US
is based on the NSFnet T3 backbone. The problem is not with having a
limit, but with having a limit based on the technology we used in 1987.
The lines have become two orders of magnitude faster, but the official
file limit is still 300k. This prevents some members of CREN and
associate networks from obtaining one of the few software packages which
is both very popular and without any Internet equivalent - in other
words, one of the few assets that BITNET still has. Is this what the CREN
board wants?
Eric
|