I think the mutt way is a reasonably good one: apply the limit when it's first commanded, but not on refreshes thereafter. Quite often I've limited to new messages, skipped past something, then decided to go back to it - and it's still displayed, even though it's no longer new.

Contrast newsbeuter, which makes a message vanish from the unread list as soon as you open it - so going back becomes a tedious dance of returning to the index, showing all messages, finding the one before the first unread, looking at it again, looking at the next one, then remembering to turn unread-only back on.

My initial thought is that makes a great simplification. I just need a persistent set of "current messages". The display code would always display those.

I'd update that set if:

• The limit in force were changed.
• The number folders currently selected changed.

I think I could implement that very very easily; so I'll have a stab at it tomorrow and see if it works out as well as it feels like it should.

Thanks very much for the idea.

This is a problem in many email clients (well at least mutt and thunderbird) and a common cause of missed mail. If a message pane is open (however minimal), scrolling though a mailbox marks messages as read.

I don't think it's a decision that should be left to the computer. I decide when I've read a mail and would prefer to mark it as read once I've done so. This would make the read flag far more reliable in human terms - ie message has been read rather than displayed, and if I tag a mail as read in a new message view I'll not be surprised when it disappears.

Thanks for making pwsafe available. I just compiled your source tree on my Loongson-2F and now I'll have it at least until the next version of Debian is out. I can't find out why they ditched it though. Do you have a link to the list of bugs which made them pull it out of stable? Again, thanks for saving my day.

The bug list is here:

• http://bugs.debian.org/cgi-bin/pkgreport.cgi?pkg=pwsafe

I'm not sure which was the main reason though..

Interesting, John, that you want to decouple the "mark read" from "reading messages". I've just switched the code from clearing the new-flag on display to doing that in the lua.

If you were to edit the default config file you could do this your own way. Something like:

function on_read_message( path )
   -- nop.  Just don't call mark-read
end

-- Mark current message as new/read
keymap['message']['N'] = 'mark_new();'
keymap['message']['R'] = 'mark_read();'

Or even something to toggle the state:

function toggle_message()
   if ( is_new() ) then
      mark_read();
   else
      mark_new()
   end
end
keymap['index']['T'] = 'toggle_message();'

Seeing things like this written down makes me pleased that lua is a good choice, and a real language is a good thing for scripting.