mail-ng
[Top] [All Lists]

Re: Operations requirement: no hop-by-hop

2004-02-13 13:00:00


I don't think multi-hop is exactly the problem. I don't really see serious
difference between calling it one-hop or multi-hop as long as actual hop is 
well defined. We'll always have remailing/forwarding servers - mailists 
is one example of that and its not going aways (I hope...). So with mailing
list this all becomes originator->origMTA->intermediateMTA(listserver)->
recMTA->receiver. There is also a question of certain networks being 
separate from each other and only acecssible through gateways. 

What I do think is that when *originator* submits messages to *origMTA* 
this MTA should decide where to route the message (preferably directly to 
endpoint), then it creates possible envelope AND checks that the destination
as per its envelope would accept the message (i.e. it transmits this envelope
and and waits for approval or does goes through some necessary authentication)
if the transmission is accepted, then rest of the message is transmitted.

This routing MTA if its not yet destination will now create its own envelope
and put the envelope it received in the tracking part of this new envelope
(call it repackaging, i.e. you received the letter, without opening it 
with its original envelope, put into new one and send it futher although 
the reality might also look more like you received the letter, took out 
its content cut out the original envelope's portion with address and put 
this all into new envelope). This all can repeat several times until it 
reaches the actual recMTA where the user will pickup this letter, but as long 
as we have the part about tranmission and authentication between MTA well 
defined we don't have a problem

On Fri, 13 Feb 2004, Paul Hoffman / IMC wrote:


This may or may not be controversial, but I propose that a 
requirement for sane operations is that mail-ng be single hop. 
Multi-hop was a good design 20 years ago when many entities were not 
online all the time, but it has also proven to be a huge hurdle in 
extending SMTP.

 From an operational standpoint, a single-hop design makes many other 
parts of the design much cleaner. It is also easier to explain to 
users.

Of course, if the receiving MTA wants to re-send a message to another 
MTA instead of writing it to the message store, that's fine. However, 
doing so is not part of the specification. That is, the protocol 
specifies just originator->origMTA->recMTA->receiver or 
originator->recMTA->receiver, but does not specify
origMTA->otherMTA->recMTA.

--Paul Hoffman, Director
--Internet Mail Consortium


-- 
William Leibzon
Elan Networks
william(_at_)elan(_dot_)net


<Prev in Thread] Current Thread [Next in Thread>