Re: [bug-mailutils] Problem with timeout interaction

[Jeff Bailey]

On Wed, 2003-10-08 at 05:05, Sergey Poznyakoff wrote:

> No, that would break the UIDL functionality:

Yuck.

> RFC1939 does not explicitly specify whether DELE marks are persistent
> across the invocations. The description of the UPDATE state in Section
> 6 only tells that POP3 must not enter UPDATE state and *remove* the
> marked messages on timeout. However it does not say that it must clear
> the deletion marks as well.

Oy.  That's a bit of a crazy interpretation, though isn't it?  I guess
the user could always issue RSET at the beginning of a session, but I've
always assumed that abnormal termination would generally leave you with
the mailbox the way it was before.

Further, in the real world case Outlook Express and Eudora both don't
cope with it.  (They terminate the connection after STAT and LIST don't
agree)

> However, we should decide if deletion marks should or should not be
> persistent. If they should, pop3d must not use attribute_set_deleted
> to mark deleted messages.

Or pop3d_rset could be modified to take a silent flag (Or split into two
functions, with pop3d_rset calling the silent one that does the work)
and get called during abnormal termination.

> By the way, even if we make it so, there remains yet another way to get
> a deleted message in a mailbox: simply add the header "Status: D" when
> sending your message :^)

> Opinions?

Resetting the deleted flags on abnormal termination would help the
popular clients here.  They would choke on it the first time, the
deleted flags would be reset and the subsequent connections would work.

It's not ideal, but I suspect it's something we have to live with.

If you agree, I'll toss this in.  Indicate if I should add a silent flag
to pop3d_rset, or if I should split the function in two.  I think I've
figured out the testsuite enough to do the other coding too.

So much to do... *sigh*

Tks,
Jeff Bailey

-- 
"I'm torn...  My kink or my country..."
 - Anika Stafford

signature.asc
Description: This is a digitally signed message part