Re: problems with the list

[Brian Reid <reid@mejac.palo-alto.ca.us>]

Here is how a computer mailing list works.

1. You send a message to the central mailbox. One sender, one recipient.
2. The central mailbox creates a new message, which has one sender (the
   list itself) and 200 recipients (the members of the list).
3. The 1-sender, 200-recipient message is placed in the outgoing mailbox.
4. When each message is placed in the outgoing mailbox, the computer
   starts delivering the mail. It does this as follows:
	4a. Deliver the mail to the first recipient
	4b. Remove that recipient from the list (check "done").
	4c. Go to 4a if there are any recipients left.

The problem is that step "4a" can take a widely varying amount of time.
This week, for example, the data link to London/UK is in very bad
shape. The link to Brazil is always in bad shape. If the link is
totally broken, then Step 4a is very fast: it just says "try again
later". But if the link is almost broken, then Step 4a is very
slow--sometimes many hours. For example, right now, there is a delivery
to the University of Guelph in Canada that has been "in progress" for 3
hours, and there is a delivery to the Technical University of Muenchen
in Germany that has been "in progress" for 5 hours. These delays are
caused by some kind of computer problem in Canada or Germany. But
meanwhile the people who are in line after Guelph and Muenchen are not
receiving their mail. 

The way that mail gets out of sequence is that often a link will get to
be so bad that its owners will just shut it down. When a link is shut
down, the processing is instantaneous. Instead of spending 5 hours
trying to deliver a message to the University of Guelph, it spends 5
seconds determining that the Guelph link is broken, and then moves to
the next one. Meanwhile the earlier message is still waiting for the
delivery that is never going to happen.

Brian Reid