Important: NANOG List Changes

Fri Nov 21 22:03:53 UTC 1997

Hi all -

We have turned on a couple of MajorDomo features to alleviate a couple of
problems we have seen recently and put in place a batch queueing mechanism
to improve the delivery speed.  Specifically...

1) WHO turned off 
-----------------
We turned off the WHO command on majordomo so folks can't find out who is
on the nanog list. In the past at least one spammer harvested nanog
addresses for individualized spam.

2) New Posting Process: nanog-post at merit.edu list
-------------------------------------------------
Starting next week, majordomo will restrict posting to the nanog list to
those subscribed to the nanog-post at merit.edu list. To subscribe to the
nanog-post list, send e-mail to majordomo at merit.edu with 

subscribe nanog-post

in the BODY.  You will receive a confirmation request, and once answered,
you will be able to post to the nanog mailing list.  This should reduce the
subscribe lists to lists problem (e.g. Marilyn Monroe) and requires only a
one-time subscription to nanog-post for those contributing content to the
group.

*You need not do anything if you don't plan on posting to the nanog list.*

3) Batch queuing of e-mail messages
-----------------------------------
We are currently piping it through bulk_mailer which forks off multiple
sendmail processes for the delivery.  This helps quite a bit.  I checked
the delivery times a while ago and the worst case I saw was about 2-3
hours.  This is not really that bad for a list of nanog's size.  There is
some additional tuning possible to improve things further (have bulk_mailer
fork additional sendmails, and tuning some of sendmail's timeouts for hosts
that are being un-responsive).  We are also looking at qmail which can
internally parallelize deliveries as is done with bulk_mailer and sendmail,
but without the overhead of having a separate processes.  Note that even
with qmail, there will also need to be a certain amount of serialization
necessary (to avoid depleting certain resources such as file descriptors)
and thus there will always be a certain amount of delay for large lists.

Hope this helps.  Thanks to Larry Blunk (ljb at merit.edu) for getting all
this stuff in place. Let me know if you have questions, problems, concerns.

Cheers -

Bill