[Asrg] is SMTP still really store-and-forward ?

[Dave CROCKER]

Tony Hansen wrote:
> I was talking with someone in the past year or so who had been traveling
> in southeast Asia. They experienced email on a totally different level:
> it was dependent on a router being brought in to the area by a shared
> traveling router. When the router was nearby, email flowed smoothly. But
> until the router came back around on its weekly journey (carried on an
> elephant, no less), the email had to be stored up before its final
> delivery. Email definitely required delay-tolerant store and forward
> processing.


Hadn't heard of that DTN example although some researchfield-monitoring 
scenarios are close.

The SMTP TRN (now ETRN) came out of our experience with CSNet's dial-up site 
access (phonenet) relaying to the Arpanet/Internet.  At one of the late-stage 
meetings that was designing SMTP, I mentioned this requirement for being 
connected occasionally and needing to pick up queued mail, as well as sending 
it.  When the remote site controlled the connection -- e.g., to save phone costs 
-- the other side would not know that it could initiate an SMTP push to the 
remote site.  Hence [E]TRN.

One of the fun example of Vint Cerf's talent was his immediately grasping the 
need to be able to explain the need for this, to funding agencies.  He was 
working at Arpa at the time and almost instantly said... submarines!

Actually, the most interesting thing is to hear that there are still some real 
work tasks for elephants.  In Thailand I'd heard that since they weren't needed 
for logging, entertaining tourists was the only thing keeping them alive...

d/

ps. a skeptic would note that your example doesn't explicitly state that the 
MTA-level is multi-hop, although I suppose there are MTAs on both sides of the 
occasional connection?
-- 

   Dave Crocker
   Brandenburg InternetWorking
   bbiw.net