According to Jay Soffian:
> Doesn't the presence of any messages in the new dir indicate "You
> have new mail." and the presence of any messages in the cur dir
> indicate "You have mail."
I wish, but the maildir manpage is clear on this:
HOW A MESSAGE IS READ
A mail reader operates as follows.
It looks through the new directory for new messages. Say
there is a new message, new/unique. The reader may freely
display the contents of new/unique, delete new/unique, or
rename new/unique as cur/unique:info.
This means a mail reader may leave new messages in new/ if it wishes.
Now if a mail reader was obliged to move all new messages to cur/unique:info
before opening or deleting a message, you could use the mtime on cur/
as an indication of when the new mail was read.
So to find (a) when mail last arrived and (b) when mail was last checked
there are the following options:
1. Use atime on new/ for (a), and mtime on tmp/ for (b)
Problem: some non-MUA programs scan new/ as well and change the
atime on new/. They should be fixed.
2. Use mtime on cur/ for (a) and mtime on tmp/ for (b)
Problem: all MUAs MUST move messages from new/ to cur/ first so
that the mtime on cur/ gets updated.
3. Scan all files in new/ and cur/ with readdir(), then stat() them all
and use the highest mtime for (a) and the highest atime for (b).
Problem: this is the traditional method but can be very disk
intensive because of all the stat()s. It is also more complicated to
code - (1) and (2) are easy to drop in in existing applications.
My original proposal was (1). Now I get the feeling that (2) might
be better, but at least mutt and qmail-pop3d do things differently -
mutt leaves messages in new/ if not read, and qmail-pop3d deletes DELEted
messages directly from new/ without ever moving them to cur/
Now another solution would be:
4. Get mtime from tmp/, cur/ and new/. the mtime on tmp is (a): last time
a new message arrived. MAX(mtime(cur), mtime(new)) defines (b): when the
maildir mailbox was last checked for new mail.
Problem: maildir(5) needs to state that a MUA MUST move the message
new/unique to cur/unique:flags unless it is deleted.
This last proposal would work with most existing applications including
qmail-pop3d and applications that scan new/ and cur/. AFAICS, only
mutt would need to be fixed - a minor issue.
Remember why I started this: to be able to make shells and status bars
etc aware of maildir without big changes. Changing the app from stat()ing
the mbox to stat()ing maildir/{cur,tmp,new} is much easier, takes less
code and is more portable then to add a complete readdir() / stat() loop
to the application.
With just one added requirement for the MUA - it MUST move the message
new/unique to cur/unique:flags unless it is deleted - this would be possible.
Besides, most MUAs (including qmail-pop3d) behave this way already.
djb?
Mike.
--
Indifference will certainly be the downfall of mankind, but who cares?
