My original question paraphrased was:
What causes the delay from the submission of a post to its
appearance in the mailbox of subscribers? The delay does not
make for an efficient turnaround on problem submissions or for
that matter postings of summaries to postpone further discussion.
****************************************************************
>From glenn@uniq.com.au Tue Aug 29 03:18 EDT 1995
Last time this came up it was noted that the machine ra.mcs.anl.gov is
a machine donated to sun-managers by that organisation. It is an old
Sun 4/110. If you'd like a faster turnaround I'm sure they'd love you
to donate a new sparc 20 or similar :-)
Remember that this group has > 50,000 subscribers and it can take along
time to explode that mail out.
Maybe if everybody donate $1.00 each then we could afford to buy a
faster machine. How about you organise it? :-)
Only 24 ( hours), how lucky you are :-) Last time I posted a question I didn't
see it for three days.
I don't think it's the volume of questions but the volume of outgoing
mail. ra takes one submission mail and turns it into tens of thousands
of outgoing mails. That's where the bottle neck is I think.
Anyway, has your question prompted many other responses? It will be
interesting to see the summary of and what the general consensus is on
improving the situation.
****************************************************************
>From pluto!perryh@qiclab.scn.rain.com Wed Aug 30 02:04 EDT 1995
I suspect it has a lot to do with the load on the main server,
ra.mcs.anl.gov. This relates to the number of addresses in the
first-level list, and the number of messages sent to the list.
One thing which might help would be to make the tree narrower (and
consequently deeper) by having more intermediate exploders, however
this would not be the easiest thing to administer.
****************************************************************
>From skidmorb@ochampus.mil Tue Aug 29 10:38 EDT 1995
My turn arounds have been as fast as 2 minutes buit never longer than 10.
check out your link....
Bruce
****************************************************************
>From tkevans@eplrx7.es.duPont.com Tue Aug 29 10:52 EDT 1995
>
You can contribute to making it better in two ways:
o Be double-damned sure that you don't post *anything* that
is not truly urgent, and that you have exhausted all other
sources of information--man pages, FAQ's, WAIS index of
summaries, newsgroups. The fewer messages that get posted,
the faster they'll all move. If you see inappropriate
postings, flame the poster (privately, of course).
o Contribute a new computer on which to host the list. Last
I heard, it was running on a 4/110.
****************************************************************
From: kalpeshd@daffy.netrex.com (Kalpesh Dharia)
Hi,
If a me too helps making it any better (I mean emphasizing that you're not
the only one who felt that way), here's one from me.
-Kalpesh
****************************************************************
>From Daniel_V._D'Errico.RochXES@xerox.com Tue Aug 29 13:45 EDT 1995
Similair mailing lists exist for SGI, DEC ultrix, osf IBM AIX, IBM PC, PC,
Mac and a few other bits and pieces of computing machinery. One is for Cray
Computers.
They all lurk on remote systems all over the internet. Happy hunting!!
****************************************************************
>From Ian_MacPhedran@engr.usask.ca Tue Aug 29 14:09 EDT 1995
Well, after having lived with UUCP and BITNet (store and forward
networks) for too long, I'm very happy with the way this service works.
However, I believe the service to be slow because it is one hop from the
central machine to my mailserver. The delays in mail seem to be (in
hours:minutes:seconds):
dyz.csph.psu.edu -> zygote.csph.psu.edu 0:00:50
zygote.csph.psu.edu -> ra.mcs.anl.gov -0:08:12 (probable clock missync)
ra.mcs.anl.gov -> engr.usask.ca 14:18:32
So the entire delay was involved on the sun managers mail handling
machine. This is likely due to its having to deliver messages to long lists
of subscribers. By spitting the subscription up into smaller "regional"
lists, the turn-around time on mail would be faster, as ra.mcs.anl.gov
would only have to worry about delivering to the distribution lists.
If you were to split the list up, then you would run into administrative
problems (where do you send mail to subscribe/unsubscribe, etc.) as well
as the problems you indicate (multi-hopped mail and down mailservers).
You also have problems getting volunteers to host this on their mailhosts.
You could try sending mail to postmaster@ra.mcs.anl.gov. That person
(those persons?) are in charge of mail for that machine, and might be
able to help you out with explaining why these delays are occuring.
As I indicate above, I don't believe this is the situation. It's just
that the mailserver is busy. (Likely it's doing other things as well as
handling this list.)
****************************************************************
>From mark@leav-156-100.army.mil Wed Aug 30 11:18 EDT 1995
I am not a Sun Managers list system adminstrator;
but I have sent several questions out.
Observation: I have noticed that I start getting responses before I get my
question posted to me from Sun Managers (sometimes even 15 hours before).
Thus some people are getting my question before I am.
I can only (in my ignorance) attribute this to the processing of
the large sun-managers mailing list.
If I remember correctly, this problem is exactly why Usenet was started:
mail is a poor distribution system for large numbers of people.
****************************************************************
>From kevin@uniq.com.au Tue Aug 29 20:36 EDT 1995
One thing to remember is that even though it may take a while for you
personally to get the distribution, others have seen it and can have
replied already.
The couple of times I've posted, I've had responses within an hour. And
it *does* take a while to spread a piece of mail to bijillions of people...
This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:10:33 CDT