Intro for LSTOWN-L subscribers: for several months I have been reporting
what I call "errors" in the ':internet.' tags of a number of nodes. As
you know these tags are used for BITNET<->Internet address mapping. While
about a dozen sites have fixed their entries, most of them have replied
that they think the tags are not in error, based on the description of
the tag in NEWTAGS DESCRIPT. Historically, the first version of the tag
description was very vague and close to being unusable, which didn't
matter since nobody was making use of the tag. When LISTSERV started
supporting it in release 1.6e, I was asked to write a better description
for the tag which is more or less what is now in NEWTAGS DESCRIPT (a
couple changes in wording were made later on to clarify the meaning).
Unfortunately I had forgotten to document the aspect of consistency
across entries (as opposed to consistency of a given entry), so the specs
do not say that the incorrect entries are incorrect and many people
refuse to change them, saying that they don't see the problem. The reason
many don't see the problem is that the tie-breaking algorithm used by
LISTSERV chanced to produce the behaviour they expected. This, however,
is totally random and may change with the addition of any node referring
to the same Internet address, unpredictably.
After a long discussion with the various people involved with updating
NEWTAGS DESCRIPT, it turns out that everyone except the BITNIC (acting on
CREN's behalf) is willing to alter the specifications, given that
LISTSERV is the only known user of the tag and that this reflects a
requirement which has been there from day 1 and which I had merely
forgotten to document, rather than something required by a change in the
software. I have decided not to make a fuss about that as I have better
things to do than fight religious battles with the BITNIC. I accept the
fact that the tags which I insist on calling incorrect are in fact
perfectly correct as per the definition of NEWTAGS DESCRIPT and that
people are free to keep them if they want. However, it has been a
permanent restriction of LISTSERV from day 1 that this produces
unpredictable results with LISTSERV, and I have invested 5 minutes
reversing the tie break algorithm used by LISTSERV to demonstrate one of
the problems these perfectly correct but non-working tags can cause. With
release 1.7f, a number of sites will find themselves without a working
:internet tag (and conversely a few sites which couldn't figure out why
theirs didn't work will suddenly find that it now works). Since LISTSERV
is the only application using this tag, this will hopefully give people
an incentive to replace their perfectly correct and acceptable but
non-working tags with other, equally correct and acceptable but also
working tags. By introducing the problem for everyone at the same time, I
can warn list owners in advance and the change will not be a surprise to
anyone; this way, both my and the list owners' time is saved. For the
record, the last time I had to investigate such a case of "non-working"
internet tag, I spent about 2h staring at the code and running all sorts
of traces for nothing because the conflict was with a node whose name was
completely different from the name of the main node, and I couldn't guess
they would have been related. To understand the problem I had to write
the program which I now use to generate the monthly trouble reports I am
posting to NODMGT-L. And that was after the list owner had hesitated for
a week to contact her local LISTSERV expert, who had also wasted time
trying to figure it out. If you don't mind, I'd like to have a working
network - whether it is legally correct or incorrect I don't care, as
long as it works.
Eric
|