Re: Any Comments on the Draft?

[Dave Cridland <dwd@xxxxxxxxxxxxxxx>]

It's occured to me that this may not actually be required.

As far as transient connections with SMTP are concerned, there are four 
scenarios.

I'll define a few terms, first:

A "client" is a machine which has the terminal SMTP server on.

A "provider" is an entity which owns a network containing a mail relay for the 
"client".

1) The client is connecting to the provider directly, and has static IP 
address.

2) The client is not connecting to the provider, but to somewhere else, but 
still with static IP.

3) Same as (1), but with dynamic IP.

4) Same as (2), but again with dynamic IP.


There are solutions for (1) and (2) in existence and heavy use, using ETRN, or 
some finger incantation, and these work well.

There is a theoretical solution for (3), by using the DNS Dynamic Update 
protocol, combined with the dialup authentication layer. It's tricky, and I 
know of no provider offering such a service, but it is, in theory, possible.


It's (4) which therefore has the problem. But I'm concerned about rewriting 
parts of the protocol in order to support this. I feel it would be of more 
benefit to create some kind of host<->host authentication seperately.

This could then be used to create DNS records enough to be able to use a 
standard ETRN command.

It should probably also be noted that (4) is grossly inefficient, and slower 
in most cases than connecting directly to the provider as in (1) or (3).


Everyone else will probably disagree, though. :-)

Dave.
-- 
Dave Cridland -- Network Systems Administrator -- Cerbernet Ltd
Job/Employer information provided purely for interest.  This is
not an official communication from Cerbernet Ltd, and any views
expressed herein are personal, as well as wrong and impolite...