ADMIN: Duplicates and related admin stuff...important

[Eric Ray <ejray -at- raycomm -dot- com>]

Gang,
Sorry about the duplicates. Here's the story:
The list server crashed between 3 and 4
am Pacific time yesterday. I paged the facility operator
and he brought it back up within 1/2 hour or so. Unfortunately,
I had to choose between possibly losing messages to the list
(by completely flushing all the messages that the server
was trying to send) or not flushing the outgoing mail and
risking duplicates. Obviously, the "keeping everything" option
won, but the duplicates were more severe than expected. 

Background: When the list server is restarted at any time,
it tries to deliver anything that it doesn't think has
been delivered. If the list server crashes (irreproducible
hardware failure like bad memory seems the most likely
culprit, with simply running out of resources as a good
second), the server usually does not get to update its records
with which messages remained undelivered to whom. That is,
if it delivered message 1 to 99% of the members but
didn't update the records as it crashed, then when it comes
back up, it'll try to redeliver all copies to everyone. If you
were one of the 99%, you'll get duplicates of message 1. 

Minor listowner tirade: My address is in every message to the
list as the contact address for problems. If there are
list problems, contact me. Do not post to the list. 
Particularly if someone else posts to the list, don't post
to the list. Why? Setting aside the irritation factor in
listowner land, each message posted to the list takes the
list server about 10-20 minutes to deliver to everyone on
the list. If we have 5 or more messages about "yup, me 
too" going to 4500 subscribers who cannot do anything
about it, that completely maxes out the server for an 
hour. If the list server is already behind (because it's 
diligently trying to make sure nobody missed anything),
those "me, too" messages make it that much further behind.
As a list reader, I'd rather see another five messages of
_content_ than another five messages of fluff. 

In related news, I'm often asked why some messages come
in far later than others posted at the same time. 
Here's the scoop:

Lyris (the list software) is optimized for speedy mail sending.
One of the side effects of that is that it's fairly impatient
when talking to various mail servers. If it contacts your
mail server and at that particular second, your mail
server is busy with other stuff, it won't wait--it'll just
go on to deliver the next message. Your message
goes into the retry in 4, 8, then 16 hour queue. 

Let me know if you have questions or need more information,
and again, sorry about the duplicates.

Eric
TECHWR-L Listowner