Re: Can we back up a bit and ask some basic questions? An alternate model

[Martin Duerst <duerst@xxxxxx>]

At 20:01 03/02/15 -0800, Paul Hoffman / IMC wrote:

> At 7:07 PM -0500 2/15/03, Martin Duerst wrote:

> > Well, I think we can distinguish the following main situations here:
> > (taking Japanese as an example)
> >
> > 1) You don't know Japanese, so you won't even try.
> > 2) You have such an address yourself, and if your MTA supports it
> >    on incoming mail, the chances are high it will work for outgoing mail.
> > 3) You have tried it for another, similar, address, and it failed,
> >    so you won't try it again unless you have good reasons to assume
> >    it might have improved.
> > 4) You'll give it a try, and see what happens. You might also
> >    cc the mail to the older (ASCII-only) address to make sure.
>
> Agree, almost. That last sentence makes an assumption that is not warranted.

You mean the assumption that an ASCII-only address is available?
That assumption is not warranted in all cases, but in most.
Initial deployment of IMAs will in most cases be of the form
'hello, I now got this new Japanese address, can you see if it
works?', and only later will first time IMA exchange (e.g.
'here is my IMA, please send me mail') increase.


> > Well, I have my 'on-the-road MTA', but I don't use its MTA.
> > SMTP after POP and now SSH tunneling work fine fine for me,
> > and wouldn't be affected. Similar for VPNs that many companies
> > use these days for security reasons. Of course, individual
> > mileage may vary.
>
> So IMAA-UTF8 is only for experts who know how to use tunneling. No, thank you.

I'm not sure there is that many people left between those who
don't travel, those who use webmail when they travel, those
who use the same ISP when they travel, and those who have
to tunnel or VPN because of company policy.
Anybody has some estimates?


> Yes, it would be quite different. Most ESMTP extensions are non-critical; 
> the message can still be delivered if the extension is not available. 
> IMAA-UTF8 would be critical: if any MTA in the path didn't support it, 
> that MTA would have to make out-of-band guesses about how to deliver the 
> message or would have to send back a "message is undeliverable" response. 
> Both are pretty bad, and I'm not sure which is worse.

Well, the assumption is still that a destination that wants to
accept IMAA-UTF8 will make sure it upgrades its firewall, too.
So as long as the firewall isn't upgraded, nobody is actually
going to use IMAs at all (except maybe internally). This may
change the rate of adoption, but it doesn't create additional
external dependencies.

Regards,    Martin.