On Fri, 20 Nov 1992 20:20:24 EST Alexander Dupuy - CS type militant
<[log in to unmask]> said:
>Well, I'm not an expert on BITNET, but I thought that there were only
>two public INTERBIT gateways on BITNET (you said there were more on
>EARN).
It is more complicated, there are a dozen such gateways in the US but
only 2 have been registered in the topology database for politico-
operational reasons (which fortunately are being worked on). While
LISTSERV only sees 2, all 12 do get to handle LISTSERV traffic; it's just
that LISTSERV may forward mail to a remote gateway when the local host
runs one because it isn't declared in the topology database. The 2
registered ones do get a lot more traffic than the others, but they have
such a big capacity that 8500 extra subscribers don't make any
difference. My initial concern was that this might fall into one of the
smaller INTERBIT's.
In fact we then learnt that UBVM uses a modified configuration file to
direct these distributions to a local SMTP/MAILER, so the load is on UB's
computers and lines.
>But you yourself pointed out that BITNET LISTSERV's idea of the best
>peer for Internet subscribers was likely to be an EARN site (not very
>optimal for lists with mostly U.S. Internet subscribers),
So don't put peers in Europe, or then tell LISTSERV not to automatically
forward subscriptions there until the US INTERBIT gang have got their act
together. How would the unix list server spread the load anyway?
>Probably because most Internet subscribers don't know very much about
>mail-based server technology.
And perhaps because they are happy with what they are using now?
>Un*x L*STSERV is relatively recent, and the interactive features have
>only become available in the latest release.
Well I've seen unix people explain the complete lack of usefulness of the
TELL interface to the VM LISTSERV hundreds of times (the argument being
that mail is just a few seconds slower and doesn't take more keystrokes
if you have an alias for your main LISTSERV). I personally find TELL very
useful, but I wouldn't ask a list owner to increase his workload just so
I could use TELL the 5-10 times a year I send a command to his server.
>And given the different natures of the BITNET and Internet
>architectures, I think it is pointless to try and find one peering
>interface which works for both of them.
Well I beg to disagree. The protocols LISTSERV are using work with the
Internet with the addition of one new primitive. At the moment LISTSERV
still requires an NJE link, but this is only due to (a) the present lack
of a mechanism to register non-NJE servers in the various databases
without breaking older servers which do not have the code to handle them
and (b) a half dozen routines, independent of the protocols, which still
assume NJE addresses. Both items are going to be addressed, it just isn't
a high priority item at the moment. FINALLY getting a mailer able to
handle lines longer than 80 and supporting it was much more important to
me.
>Since most users don't ever see the peering interface, I don't see this
>as a problem.
If you assume the BITNET and Internet servers shouldn't be able to peer
with each other, you are right.
>Which is what is happening, especially with other Unix mail-based
>subscription managers. I find it ironic that Tasos and I are the ones
>arguing for as much subscriber-interface compatibility as possible
>between the BITNET and Internet worlds, while you and most of the rest
>of the Internet list-managers seem content to ignore each other's work,
>rather than trying to create a more powerful synthesis of the two (and
>no, this doesn't mean I want to force everybody's mail servers to be
>running the same implementation).
First, let me point out that by no means do Tasos and you have the
exclusivity of worrying about syntax compatibility. I find this statement
and the use of the word 'two' when referring to BITNET and Internet list
managers a bit presumptuous, and will refer you to the 1.7e release notes
for the name of a competing product whose authors are concerned about
this sort of things and which, in my personal opinion, is superior to
what you are using.
Second, I really don't understand what I am accused of not having done.
Joe decides that it is Bad to only have evil IBM stone-age list servers
with brain-damaged syntaxes and makes a 200-liner in perl which uses
'add' rather than 'subscribe' and 'scan' for 'review' or whatever. What
*am* I supposed to do about it? Waste time telling Joe it is uncool to
have incompatible syntax when I saw his announcement saying "more
sensible command set", and be called a fascist for attempting to impose
my syntax on him? Change LISTSERV to do like Joe? Let's be real! Granted,
the present situation is that there are only a handful of "leading" unix
list servers with a non-negligible user base, but this whole thing
started as a sudden surge of 200-liners here and there, each of them with
different and incompatible syntaxes. What was I supposed to do, whine? I
suppose that is not relevant and we should look forward. Fine, so we have
a handful of leading unix servers which we should worry about and a
couple score 200-liners which we can ignore. The handful have different
and incompatible syntaxes and a large enough user base to make changes
difficult, and I respect that. Now, my user base is probably an order of
magnitude larger than the cumulated user bases of all the unix servers,
so who should change his syntax? I have made upwards-compatible changes I
felt I could make to help the end users. I am open to more suggestions.
But I simply can't make incompatible changes, and this is not open to
discussion.
Eric
|
