Re: [Nmh-workers] XOAUTH2 integration, and a few questions

[Ken Hornstein]

>Well, you can (and I did in sendfrom.c) by relying on the fact that (most)
>MH switches only use their last value if repeated.

I'm just thinking about a hypothetical world where there would be two
different mail servers that support XOAUTH; you don't get to make the
choice in postproc, because send(1) is the guy who generates the bearer
token.

>There'd be 6 or 7 of them, so this is getting a bit ugly for switches, but OK.
>But should we consider alternatives, such as pseudoheaders?

Hm.  That seems like it's wrong.  Just from my feelings about things,
headers in the message are attributes _of the message_; this seems more
like something that is just IPC between send and post.  Also, it would
be hard to modify this in a postproc.  I mean, it could be done, but it
just seems cumbersome.

I looked at prior art; besides -credentials, which you added, there
is also -idanno, which is the number of the file descriptor to write
annotations to.  It seems like this is the best fit, since it my
mind it falls under "stuff send(1) wants to communicate to post(8)".
Environment variables are another option (these are just the profile
entries, not the bearer token itself), but those might be cumbersome as
well.

This brings up a related point; should we document all of those internal
switches like -idanno?  I would say yes.

>Or let me know if you want me to do it.  The OAUTH code is quite clean so
>this won't be hard.  And I need to integrate sendfrom into send, and it
>will have to change a bit.

I'm fine with doing it; spread out the work and all that.  Agreed that
it looks pretty straightforward.

>And it would be good to have Eric's thoughts.

Definitely!

--Ken