spam2.el again (was: bug#9434: [nnir] error quitting group)

[Ted Zlatanov <tzz@lifelogs.com>]

Following up with a changed subject because the original subject was
completely wrong.

Ted

On Tue, 24 Jan 2012 15:49:15 -0600 Ted Zlatanov <tzz@lifelogs.com> wrote: 

TZ> On Sun, 04 Sep 2011 18:44:08 -0800 Dave Abrahams <dave@boostpro.com> wrote: 
>>> This is definitely NOT the case for nnir right now (but will be as
>>> soon as I push my re-write). So I might be wrong about the
>>> spam-related variables being global, and I'm sorry I suggested
>>> it. Maybe Ted can set me straight?

DA> I sure hope it's supported.

TZ> The spam.el code is not buffer-localized.  All the variables it creates
TZ> are global.  I'd rather rewrite the whole thing as spam2.el than try to
TZ> fix this one thing, because there are so many other issues with spam.el:

TZ> - hard to configure

TZ> - not loaded by default

TZ> - not well integrated with the Gnus registry

TZ> - old and unused spam/ham processors

TZ> - copy/move logic needs to be much cleaner

TZ> - general logic is hard to follow, I was learning ELisp (I still am, but
TZ>   hey, it's only been 15 years)

TZ> - elegant original design by Francois Pinard corrupted by my primitive
TZ>   knowledge of ELisp ;)

TZ> - many new ideas floating in my head, especially "jam" handling (spam
TZ>   you have legitimately requested)

TZ> - at least 2 old threads had improvement suggestions for spam2.el, and
TZ>   I've saved them for a rainy day

TZ> CC to ding mailing list for comments.