[ih] Design choices in SMTP

[Dave Crocker]

On 2/7/2023 10:37 AM, Guy Almes via Internet-history wrote:
>   So nowadays, all those round trips really matter, while in the 1980s 
> they more or less didn't. 

They matter sometimes.  Other times, not so much.

Their relative effect on the time needed to complete a single, overall 
exchange? Sure. In absolute terms, it rasies the unit cost.

The end-to-end effect on UX, for author and recipient of email? Nah. One 
of the nice things about email is how much cruft can be tolerated, as 
long as it is in the background.

What about for real-time chat?  I bet nah to that, too, as long as 
overall server performance remains good.  (See below.)

Real-time video?  Probably a big yah.

And an orthogonal issue is overall systems effects, at scale. When 
servicing gillions of user mail on the same server, does the aggregate 
of all those micro-interactions affect system performance?  My 
understanding is yah, again.  Not necessarily obviously, having to hold 
onto the message longer messes with queue sizes, I believe.

d/

ps. For obvious reasons, geosync paths alter the real-time UX calculus 
fundamentally.  Extra round-trips a problematic.

pps. Going from SMTP to ESMTP was accomplished with a design process 
that paid quite a lot of attention to keeping the micro-transaction 
count the same.

-- 
Dave Crocker
Brandenburg InternetWorking
bbiw.net
mast:@dcrocker at mastodon.social