I was being ironical of course. This is just unfeasible, and it is not
desirable either. Nobody would be willing to donate enough CPU time (on a FAST
CPU) and DASD space for this. Note that they would also have to make sure the
server operates 24 hours, that the computer is never stopped for maintenance,
that the link is always up, etc. They would have to allow just anybody to
create any list for whatever purpose they might want. Thus if one of my users
wants to create a list to discuss whether Marilyn Monroe's most pretty devices
was her legs or her eyes, they'd have to accept it. That is, they'd have to be
politically free, no administrative constraint, etc. They would also have to
be available all the time so that you don't wait as long for a new list
creation than you have to wait for a reply from [log in to unmask]
As each archiving request would be directed to this only server, the node
housing it would have to purchase a dozen 50kb lines to handle the traffic.
That is, yes, people from France would send their requests to the USA since
there would be no place near France to ask for the files.
The output of a LIST LONG command would be around 7,500 lines or so. It's
already more than 2,500 at FINHUTC, and we'd have at least thrice the number
of lists that we have there. A mere LIST command might return as many lines of
output as a $DU command sent to FRMOP22 (don't advise trying it), and would
take 1 minutes of CPU (my 4381-11 units) to process.
In case you're not aware of the load the major network servers are subjected
to: I went to DEARN in Easter to install the RFC822 stuff on its NETSERV for a
real-scale test. NETSERV there is working continuously. That is, it kept
receiving requests about the same speed it could process them. It took more
than 5 minutes before we could stop it. And that is only for Germany. Okay,
LISTSERV is faster than NETSERV but still... Building up a DISTRIBUTE job for
the 300+ recipients of a large list like MAILBOOK would take around 60 seconds
of 4381-11 processor time. That is, requests would be coming in faster than
they could be processed on my machine. A 3090 could easily do it but LISTSERV
is not what it's been purchased for, huh?
When the PEERS FILELIST stuff is operational, it will be possible for each
server to know the locations where the major network-wide lists are housed,
and to forward SUBSCRIBE requests to the appropriate server. And I think
that's the way we should go. Most of this headache would not exist if we had a
serious LISTSERV GROUPS file...
Eric
|