nnmaildir & expiry (again..)

nnmaildir & expiry (again..)

[Raphaël Berbain]

I already asked this question on gnu.emacs.gnus[1] some time ago with
no luck, so I hope someone around here can enlighten me.

I am having problems setting up expiry to work with nnmaildir:  I use
auto-expiry. Expiry age and target settings work correctly.  The
problem is, after being expired the article appears unread again in
the target group.  The behavior was different with nnimap, where it
was marked read in the target group.

What should be the right behavior ?  It looks to me that marking as
read would make more sense.  Regardless of this, I think both backends
should behave in the same way.  And last, if articles shouldn't be
marked as read by default, is there a way to change the behavior ?

Footnotes: 
[1] Message-ID: <86pt68p6ko.fsf@ID-110038.user.uni-berlin.de>

Re: nnmaildir & expiry (again..)

[Paul Jarc]

raphael.berbain@free.fr (Raphaël Berbain) wrote:
> The problem is, after being expired the article appears unread again
> in the target group.

If you want all the articles in the target group to be marked as read,
you could set (always-marks . ['(read)]) in the group parameters.

> What should be the right behavior ?

Copying all the marks (except 'expire) from the original article would
probably be good.  I'll have a look at that.

> Regardless of this, I think both backends should behave in the same
> way.

Yes, the backend interface ought to be redesigned to put less (i.e.,
no) policy decisions in the backends.  But that's a big job.


paul

Re: nnmaildir & expiry (again..)

[Raphaël Berbain]

prj@po.cwru.edu (Paul Jarc) writes:

> raphael.berbain@free.fr (Raphaël Berbain) wrote:
>> The problem is, after being expired the article appears unread again
>> in the target group.
>
> If you want all the articles in the target group to be marked as read,
> you could set (always-marks . ['(read)]) in the group parameters.

I have thought about that, but the doc says it's an experimental
feature and likely to change ; Also AFAICT it doesn't alter the actual
stored-on-disk marks ;  And last but not least, I do not want all
articles in those groups to be read.  Well, not for all concerned
groups anyways.  it's a good work-around for some of them, however.

>> What should be the right behavior ?
>
> Copying all the marks (except 'expire) from the original article would
> probably be good.

Agreed.

>> Regardless of this, I think both backends should behave in the same
>> way.
>
> Yes, the backend interface ought to be redesigned to put less (i.e.,
> no) policy decisions in the backends.  But that's a big job.

Agreed too.

While I have your attention (warning, subject hijacking + feature
request coming..), nnmaildir stores marks quite unusually compared to
other maildir-aware MUA.  It uses files in
<maildir>/.nnmaildir/marks/<mark>/file.  Others tend to use the info
part of the message filename.

What is the rationale behind this design ?  The only reason I could
think of is that Gnus has more marks than some other MUAs/imap
servers, perhaps with some mark semantics confusion kicking in.
However, it would be useful to me to get nnmaildir to handle marks
this way as well (at least the 'read' mark) because my maildirs are
also served through imap.  I think I once heard about a script (perl,
IIRC) that did some kind of sync between both systems, but I can't
find it again.

Re: nnmaildir & expiry (again..)

[Paul Jarc]

raphael.berbain@free.fr (Raphaël Berbain) wrote:
> nnmaildir stores marks quite unusually compared to other
> maildir-aware MUA.  It uses files in
> <maildir>/.nnmaildir/marks/<mark>/file.  Others tend to use the info
> part of the message filename.

Yes.  Having marks as separate files is more concurrency-friendly, for
programs that all use this scheme (right now, that just means
concurrent Gnusae): when one program alters one piece of information
(say, a mark), it does not make it harder for another program to find
another piece of information (say, the message contents).  I posted
this idea to one of DJB's lists, but didn't get much response.

> The only reason I could think of is that Gnus has more marks than
> some other MUAs/imap servers

There's that too.

> I think I once heard about a script (perl, IIRC) that did some kind
> of sync between both systems, but I can't find it again.

<URL:http://multivac.cwru.edu./prj/marksync>


paul

Re: nnmaildir & expiry (again..)

[Raphaël Berbain]

prj@po.cwru.edu (Paul Jarc) writes:

>> I think I once heard about a script (perl, IIRC) that did some kind
>> of sync between both systems, but I can't find it again.
>
> <URL:http://multivac.cwru.edu./prj/marksync>

*blush* Thanks.