[Top] [All Lists]

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

2013-12-12 23:19:16
As with my attach addition which is way more contentious now than it
was when I implemented it, my proposal on side is compatible with the
existing ui.  It would look the same if you didn't turn on any of the
"new" options.  Could even leave mhlist et. al. in place for any who
love 'em.  It's true to the "don't break things" philosophy.

So, out of curiousity I went back and dug into these changes ... I see
they were committed on August 19th, 2002.  More than 11 years ago!  The
previous changes to that were all in July, and then a HUGE gap from
those to March of 2001.  That was also back during our transition from
mhost to savannah ... and from the mailing list archives, it doesn't
look like you sent out any email about the change either (correction:
looks like there was a brief announcement on 2002-11-29).  So ... I
expect the reason that it wasn't controversial was that no one noticed
(I actually didn't notice that support for years, I am sure).

This gets into larger issues in terms of code changes, long term direction,
etc etc.  Since we've touched on these meta-issues, let me speak to them
more directly.

Over the years we've been slowly limping toward getting better MIME support,
but it's mostly been via what can kindly be called "hacks".  E.g., the
attach support, replyfilter (and associated framework), and mhfixmsg.
Most of this code went in with little discussion; well, there was some
discussion about mhfixmsg, but I rammed in the framework for replyfilter
through without any discussion (I think people were just happy to have
some solution that they didn't care; the only serious discussion was where
to put the replyfilter script).

These stealth hacks are fine in and of themselves; they're driven by
real needs, and more code is always welcome.  But we're getting to the
point where all of the easy stuff is done, and we're left with harder
and harder things.  Those things require us to really think about some
of the larger issues that we need to grapple with.

This is why I brought up the issues regarding always running mhbuild; I
thought it was simple, but it turns out it ain't (like many things in
nmh).  That gets into the larger meta-issue of "How do we want people
to compose MIME content, anyway?".  We're a little bit closer to making
that better now.  I'm not always sure of the right answer; that's why I
punt a lot of these things to the list.  It also forces me to defend my
ideas, and frequently fighting for them makes them better (I think my
new idea for mhbuild/attach is better than the old one).

I guess my point is that I think the time of stealth hacks, especially
when it comes to MIME support, is over.  We need to start thinking
larger and long-term and grapple with serious issues that have been
languishing for a few decades.  To me that's the core of all of these
discussions.  I think we're working them out, it's just going to take
a while.  That's not to say that everything needs to go through the
mailing list, or that we need to implement something like Gerrit; use
your best judgement.

I suspect like Mike O'Dell suggested a little while ago we're eventually
going to have to have a MIME multiplexor at some point to really handle
everything properly.  How that will work or be implemented really is up
in the air, though.


Nmh-workers mailing list

<Prev in Thread] Current Thread [Next in Thread>