[geeks] ADMINISTRIVIA: Changes to mail delivery policies

[Greg A. Woods]

[ On Tuesday, March 12, 2002 at 19:09:04 (-0600), Bill Bradford wrote: ]
> Subject: [geeks] ADMINISTRIVIA: Changes to mail delivery policies
>
> 1.  If my end cant contact a host to hand a mail off to (either a primary/
>     secondary MX, or the desired host itself), it will drop it, instead of
>     trying for four days.  In other words, messages expire after 24 hours if
>     they're not delivered in that time frame.  Four days was fine back in the
>     era of UUCP and non-24-7 connections, but this is 2002, and if you dont 
>     have a fulltime net connection, you should at least have a primary or 
>     secondary MX host that does.

I just want to point out that your choice of 24 hours is very bogus on
at least two fronts, and also tell you I'm not the only one saying so.

Large parts of the Internet can be easily and regularly cut off from the
rest of the net for even a full long weekend.  All the net is not in the
USA (and even there the number of back-water leaf nodes on the net is
very large.  "All our base do NOT belong to you!"

Secondary MX hosts are very rare and increasingly so.  It is impossible
to run a secure MTA with a secondary MX that you do not control, so even
if people do have secondary MXers then they will likely both become
unreachable when connectivity fails since they'll both be sitting
topologically side by side on the very same logical IP network.

Your decision is not just contrary to good common sense given these very
real circumstances -- it's also contrary to the guidelines in the newly
proposed SMTP standard as documented in RFC 2821:

   Retries continue until the message is transmitted or the sender gives
   up; the give-up time generally needs to be at least 4-5 days.

(In RFC speak that's the equivalent of a _very_ strong "SHOULD".)

Personally I find 5 days to be an absolute minimum, and seven days is
far more sane, even and especially in today's very diverse Internet.

> Yes, I'm turning into a mail BOFH.  If you dont like it, you can unsubscribe,
> or start paying for my bandwidth.  8-)

The retries (normally just three tiny SYN packets each) for 5 days
backlog for a few unreachable sites are not going to even be a tiny
pixel on your bandwidth graph, and I'm sure you won't even notice the
disk space and other sundry resources they may take up either.

-- 
								Greg A. Woods

+1 416 218-0098;  <gwoods at acm.org>;  <g.a.woods at ieee.org>;  <woods at robohack.ca>
Planix, Inc. <woods at planix.com>; VE3TCP; Secrets of the Weird <woods at weird.com>