Re: [Nmh-workers] repl doesn't like return address

[Ken Hornstein]

>    name-addr     =  [display-name] angle-addr
>    display-name  =  phrase
>    phrase        =  1*word / obs-phrase
>    obs-phrase    =  word *(word / "." / CFWS)
>    word          =  atom / quoted-string
>    atom          =  [CFWS] 1*atext [CFWS]
>    atext         =  # One of A-Z, a-z, 0-9, or "!#$%&'*+-/=?^_`{|}~".
>
>Given atext doesn't include a `.', the one or more words that make up
>phrase can't be `Dr. Seuss', or `foo.bar'.  But they are allowed by
>obs-phrase as long as they're not at the beginning.  So those two
>examples are both parsed as obs-phrase with `.' being the middle token.

Oh-ho!  I've been out-RFC lawyered!  Yeah, it seems like that is completely
valid; I missed that subtlty regarding obs-phrase.

>Ken, what's the parsing difficulty?   4.1 hints at it.  Does nmh have
>extra difficulty because it takes email address without a `@' and domain
>part?  I'm struggling to come up with examples, perhaps because I'm
>allowing too much look ahead, e.g. `a b c d e . f' takes a few atoms
>before you know if it's a phrase or an obs-phrase.

It's just .... ugh.  Have you looked at the address parser?  I hate going
in there.  But it seems we're clearly in the wrong and we need to fix it.
I mean, yeah, we should fix it anyway because literally every other frickin'
MUA on the planet handles it even if it's wrong, but this is added fuel
to the fire.

--Ken