Re: [ietf-smtp] ALPN

[Ned Freed <[e-mail]>]

> On Thu, Jul 29, 2021 at 02:57:12PM -0400, John Levine wrote:

> > >Since there are already IDs for IMAP and POP3, it seems we are down to
> > >3 (+SMTP) or 4 (+SUBMIT) unless we want to break existing applications?

> I'd prefer to see "SUBMIT" as a separate ALPN value from "SMTP".

As would I. A clean separation is always a good idea.

> > Given that clients often don't know whether they're talking to an SMTP or a submission
> > server,

> In fact they almost always do, because "SUBMIT" generally requires
> authentication, and SMTP relay does not, and the separate ports also
> help.

> > and servers don't know which one they are or maybe both, one ALPN would be plenty.

> Submission servers are explicitly configured to accept outbound mail,
> and it has been best practice for quite some time to not implement
> submission on port 25, leaving that to 587, and more recently also 465.

> > It's still pretty common for a server to offer SMTP on port 25, which
> > turns into submission if the client authenticates.

> This is a no longer recommended and diminishing practice.

>     - An MUA can always use the "SUBMIT" ALPN value.

>     - An occasional MTA that is configured to relay outbound mail
>       via a submission server, and has been updated to use ALPN,
>       can be configured to use "SUBMIT" when appropriate.

I also don't see how this practice, which I agree is diminishing, is especially
relevant to ALPN assignment. If you really want to do this in an ALPN context -
I have no idea why - you can always use the SMTP ALPN to do it.

				Ned