Re: [VM] Threading using subject

[Uday Reddy]

Goran Uddeborg writes:

> On 2012-05-12 Uday Reddy <address@hidden> wrote:
> > So, the question is, should we change the default value of
> > `vm-thread-using-subject' to nil?
> 
> Would just changing the default help?  As long as the option is there,
> I guess the code has to support it.

Yes, the code will continue to support it.  But I could declare that
threading using subject does not work correctly, and people will need to use
it at their own risk.  (I know that that warning is always on the box, but
that is more of a legalspeak.)

I might also impose some pain on the users, by pointing out when things go
wrong so that they can take care to make sure that things are working
properly before they do something critical, like deleting a thread or
forwarding a thread.  I use an integrity checker for the threads-database in
my daily use so that I know when things go wrong.  But the checker is turned
off for normal users at present.

My guess is that the need for thread-using-subject only arises these days
when we read newsgroups or mailing lists, whose servers often munge the mail
headers.  Reference threads get broken as a result and subject threading can
be used to still get back some coherence.  For normal email usage, on the
other hand, subject threading is rarely needed (unless you correspond
regularly with somebody that has a bad email client.)

> It is very rare for threading using subject to give any false hits, in
> my experience.  So I don't think it hurts.

Here are a couple of things that worry me.  Assume that the threads database
is wrong, i.e., it includes in a thread a message that doesn't really belong
in the thread.  Then,

1. If you delete a thread, then you might end up losing the message that you
you never intended to delete.

2. If you forward a thread to somebody, you might expose information that
should not have been exposed.

Ok, so that gives me an idea.  The thread operations are where unsafety
might arise.  So, I could "firewall" the thread operations so that they
don't do wrong things even if the threads database is inconsistent.  I will
try and implement that.

Cheers,
Uday