On Mon, Sep 20, 1999 at 05:12:22PM -0700, Racer X wrote:
> > > what's wrong with using the info field, in particular, info starting
> with
> > > "1," (experimental semantics)? if you don't like that, ask him to
> register
> > > another info field.
> >
> > I fail to see how this can be used to implement discrete folders.
>
> if it starts with "2," and it has an R after that, it's been read. if it
> starts with "1," and has an F after that, then immediately following the F
> is the folder name. no, this is not perfect, it's just what i thought of
> off the top of my head.
This is a creative solution that differs from the two creative solutions
already implemented in different UAs. Now there's at three ways that have
been discussed, which is more complex, not less.
> if you're talking about physically putting the files in different
> directories, i say that it's a UA problem to deal with, not a Maildir
> problem. don't blame the Maildir spec for not solving a problem it wasn't
> meant to solve.
It's really not about whether or not Maildir can be used to solve the folder
problem, and there's really no blame anyone is trying to place. We know
that a Maildir is a useful, safe place to deliver mail even simultaneously
with UA access and simultaneously with other deliveries (over NFS, without
locks).
We know that the Maildir format provides specifically for the above.
We're simply trying to establish a standard semantic for applying Maildir
to implement a common UA feature in a way that promotes interoperability
between Maildir-using UAs.
Without such a recommended standard semantic, authors will select something
they think is "nicest", which will be different from what other authors
select as being "nicest". Of course, all these authors will be right to
implement something that doesn't break Maildirs as useful, safe places to
deliver mail even simultaneously with UA access et cetera. Implementations
that break Maildirs' given functions will be broken implementations.
It's really just a matter of selecting one method that people can point to
as the preferred way. If the creator of the Maildir format would comment as
to the way he would suggest, then actually putting that method in place for
specific UAs could be done with confidence of compatibility with other UAs.
Er, ones whose authors listen to DJB's suggestion. Additionally, broken UAs
could be pointed at with confidence as implementing a broken version of
Maildir folders/subfolders.
If subfolders weren't part of some UA's folder semantics, I'm pretty sure
this wouldn't be at issue - we could use ~/Maildir/, ~/Maildir-foo/ etc and
be done. With subfolders, the ambiguity of whether alternative folders to
INBOX should be placed inside ~/Maildir/ or in ~/Maildir-foo/ is at issue.
Plus the second question of where subfolders should live. Of course, all
proposed methods _work_. But what single method should be used as standard?
Still hoping that the group can arrive at a single solution, perhaps with
Dan's assistance...
Randy
