Re: [Nmh-workers] Conflict between "mime" command and attach

[David Levine]

> >I think it'd be worth trying to solve that.  Could we look for
> >false mhbuild directives and escape them if not already escaped?
> 
> I think it would be challenging; the code that does that is the regular
> MIME parsing code.  Also, the syntax is that if it's not something that
> is recognized after the "#", then it's a content-type.

Challenging but I think it would be possible.

> >It's not the pseudo-header, it's the mhbuild directives.  "attach"
> >users can't view/modify them now.  This would be a benefit, and
> >though I have cured myself of the urge to look at them, I would if
> >I could.
> 
> Well, that's what I was thinking with #attach; that would give you the
> benefits of attach (picking the MIME type and automatically adding
> the right disposition).

Let me add that being able to view/modify the mhbuild directives
inserted by attach would be a very, very small benefit.  I don't
even think about it any more.

> >If we're going to address that, I think we should step back and
> >consider whether some other approach would make sense.  I don't
> >think adding more directive types just to simplify the user
> >interface is a good idea.
> 
> Well, I'm not completely wedded to this idea; that's why I asked for
> feedback.  What do you think we should do instead?

I just realized that I misunderstood something:  "-nodirectives"
means "don't process directives", not something else that I thought.
So, would this work:

In WhatNow?

- If you run "mime", run mhbuild on the draft.
- If you "attach", add the appropriate mhbuild directive.  Refuse if
  if there is a MIME-Version header.  Optionally, run "mime" when
  you're done attaching if you want to see/edit what mhbuild did.

In send(1):

- If there were any attach calls (Nmh-Attachment header), run "mime"
  if there isn't a MIME-Version header.

In post(8):

- Run "mhbuild -auto -nodirectives".

To deal with the false directives when using attach:
WhatNow could look for any lines that begin with # during
the first call to "attach", and warn or refuse.  Those who
use # for replied-to text would have to find something else.
Or we could get more sophisticated, back to the "challenging"
aspect above.

This is where we want to end up, right:

  1) Run mhbuild -directives if the draft has directives and
     doesn't have false directives.

  2) Always run mhbuild -auto -nodirectives.

> >At the cost of arcane-ness, which would increase the likelihood
> >of confusion of text with an mhbuild directive.
> 
> More arcane than the existing syntax? :-)

My point was that #attach is less arcane, I didn't say that clearly.
It's therefore more likely to get confused with text that isn't
intended to be a directive.

David