nnimap danger! Do not use!

nnimap danger! Do not use!

[Lars Magne Ingebrigtsen]

I just discovered that all my IMAP messages that are older than three
hours are now marked as %Seen.  Oops.

I'm investigating what's going on, but don't use Gnus at the moment if
you value your IMAP marks...

-- 
(domestic pets only, the antidote for overdose, milk.)
  larsi@gnus.org * Lars Magne Ingebrigtsen

Re: nnimap danger! Do not use!

[Steinar Bang]

>>>>> Lars Magne Ingebrigtsen <larsi@gnus.org>:

> I just discovered that all my IMAP messages that are older than three
> hours are now marked as %Seen.  Oops.

Sounds similar to my problem...?

> I'm investigating what's going on, but don't use Gnus at the moment if
> you value your IMAP marks...

Right!

Re: nnimap danger! Do not use!

[Lars Magne Ingebrigtsen]

Steinar Bang <sb@dod.no> writes:

>> I just discovered that all my IMAP messages that are older than three
>> hours are now marked as %Seen.  Oops.
>
> Sounds similar to my problem...?

Quite possibly.

But I don't understand at all what could be causing this.  This seems to
have affected groups that I've never entered.  But there are no marks
updates anywhere unless you enter a group, is there?

-- 
(domestic pets only, the antidote for overdose, milk.)
  larsi@gnus.org * Lars Magne Ingebrigtsen

Re: nnimap danger! Do not use!

[Richard Riley]

Lars Magne Ingebrigtsen <larsi@gnus.org> writes:

> I just discovered that all my IMAP messages that are older than three
> hours are now marked as %Seen.  Oops.
>
> I'm investigating what's going on, but don't use Gnus at the moment if
> you value your IMAP marks...


sob ;)

(reading this using latest gnus imap ...)

Re: nnimap danger! Do not use!

[Steinar Bang]

>>>>> Richard Riley <rileyrg@googlemail.com>:

> Lars Magne Ingebrigtsen <larsi@gnus.org> writes:
>> I just discovered that all my IMAP messages that are older than three
>> hours are now marked as %Seen.  Oops.

>> I'm investigating what's going on, but don't use Gnus at the moment if
>> you value your IMAP marks...


> sob ;)

> (reading this using latest gnus imap ...)

Yah... I'm back on a September 5 gnus, and hoping things aren't too
messed up and too many of my marks unmarked..

Re: nnimap danger! Do not use!

[Lars Magne Ingebrigtsen]

Lars Magne Ingebrigtsen <larsi@gnus.org> writes:

> But I don't understand at all what could be causing this.  This seems to
> have affected groups that I've never entered.  But there are no marks
> updates anywhere unless you enter a group, is there?

Ok, I think I may have an inkling about where this is happening.  If the
active data returned by nnimap is, say, (200 . 230), then Gnus will send
"mark as read 1:199".  So a buggy active range has catastrophic
consequences on the IMAP server.

-- 
(domestic pets only, the antidote for overdose, milk.)
  larsi@gnus.org * Lars Magne Ingebrigtsen

Re: nnimap danger! Do not use!

[Richard Riley]

Steinar Bang <sb@dod.no> writes:

>>>>>> Richard Riley <rileyrg@googlemail.com>:
>
>> Lars Magne Ingebrigtsen <larsi@gnus.org> writes:
>>> I just discovered that all my IMAP messages that are older than three
>>> hours are now marked as %Seen.  Oops.
>
>>> I'm investigating what's going on, but don't use Gnus at the moment if
>>> you value your IMAP marks...
>
>> sob ;)
>
>> (reading this using latest gnus imap ...)
>
> Yah... I'm back on a September 5 gnus, and hoping things aren't too
> messed up and too many of my marks unmarked..
>

Its been a nightmare for me recently. Its why I was whining so much
about defaulting to local server and stuff I guess. I spent two days
trying to get splitting working as it was before only to discover it was
a bug after all despite bringing it up before hand ;) Honestly, my
emails are HOT from so much shuttling back and forward between dovecot,
nnml and gmail imap via offlineimap. Still, all good fun and worth it in
the end.

Re: nnimap danger! Do not use!

[Lars Magne Ingebrigtsen]

Steinar Bang <sb@dod.no> writes:

> Yah... I'm back on a September 5 gnus, and hoping things aren't too
> messed up and too many of my marks unmarked..

Do you have an old .newsrc.eld file from before today?  If you, could
you send the section that deals with one of the affected nnimap groups
from the old file and the new file?

I'm still not understanding what's going on here...

-- 
(domestic pets only, the antidote for overdose, milk.)
  larsi@gnus.org * Lars Magne Ingebrigtsen

Re: nnimap danger! Do not use!

[Steinar Bang]

>>>>> Lars Magne Ingebrigtsen <larsi@gnus.org>:

> Steinar Bang <sb@dod.no> writes:
>> Yah... I'm back on a September 5 gnus, and hoping things aren't too
>> messed up and too many of my marks unmarked..

> Do you have an old .newsrc.eld file from before today?  If you, could
> you send the section that deals with one of the affected nnimap groups
> from the old file and the new file?

How do I get to that part?  `G E'?

Re: nnimap danger! Do not use!

[Lars Magne Ingebrigtsen]

Steinar Bang <sb@dod.no> writes:

> How do I get to that part?  `G E'?

For the new version -- yes.

In the old file, just look in the file and search for the sexp that
contains the group name.

-- 
(domestic pets only, the antidote for overdose, milk.)
  larsi@gnus.org * Lars Magne Ingebrigtsen

Re: nnimap danger! Do not use!

[Steinar Bang]

>>>>> Steinar Bang <sb@dod.no>:

>>>>> Lars Magne Ingebrigtsen <larsi@gnus.org>:
>> Steinar Bang <sb@dod.no> writes:
>>> Yah... I'm back on a September 5 gnus, and hoping things aren't too
>>> messed up and too many of my marks unmarked..

>> Do you have an old .newsrc.eld file from before today?  If you, could
>> you send the section that deals with one of the affected nnimap groups
>> from the old file and the new file?

> How do I get to that part?  `G E'?

Silly suggestion: that would mean loading the old file in gnus, and
maybe corrupting it.

I've sent the info for before and after for one group to you in email.

What I did was search for the group name in .newsrc.eld, then go to the
starting paranthesis, and C-SPC followed by C-M-f and M-w.

Re: nnimap danger! Do not use!

[Steinar Bang]

>>>>> Lars Magne Ingebrigtsen <larsi@gnus.org>:

> Steinar Bang <sb@dod.no> writes:
>> How do I get to that part?  `G E'?

> For the new version -- yes.

> In the old file, just look in the file and search for the sexp that
> contains the group name.

Right.  Actually what I did for both of them.

Re: nnimap danger! Do not use!

[Steinar Bang]

>>>>> Steinar Bang <sb@dod.no>:

> I've sent the info for before and after for one group to you in email.

Just hope you haven't started reading your real email through IMAP...? :-)

Re: nnimap danger! Do not use!

[Lars Magne Ingebrigtsen]

I think Steinar's and my problems are unrelated.  Steinar's are probably
related to the new QRESYNC code, and means that some new marks that are
set aren't synced.  I think I may have fixed (some of) it now.

I can't find any explanation for the \Seen-marks all over my groups.  I
have no idea when it happened, so it may have been something that
happened while I was testing, and it's OK now...

Could somebody who has a non-QRESYNC setup, and who doesn't care whether
marks get wiped, try the code now?  :-)

-- 
(domestic pets only, the antidote for overdose, milk.)
  larsi@gnus.org * Lars Magne Ingebrigtsen

Re: nnimap danger! Do not use!

[Rémi Letot]

Lars Magne Ingebrigtsen <larsi@gnus.org> writes:

> I think Steinar's and my problems are unrelated.  Steinar's are probably
> related to the new QRESYNC code, and means that some new marks that are
> set aren't synced.  I think I may have fixed (some of) it now.
>
> I can't find any explanation for the \Seen-marks all over my groups.  I
> have no idea when it happened, so it may have been something that
> happened while I was testing, and it's OK now...
>
> Could somebody who has a non-QRESYNC setup, and who doesn't care whether
> marks get wiped, try the code now?  :-)

If \Seen-marks are equivallent to read, I just subscribed a non critical
group with old read and unread articles, did g, did M-g, entered it, all
of these several times, and didn't see any modification in the marks.

My server is an old Courrier, which I guess doesn't have QRESYNC ?

Oh, pulled gnus minutes ago to get latest code.

HTH,
-- 
Rémi

Re: nnimap danger! Do not use!

[Lars Magne Ingebrigtsen]

hobbes@poukram.net (Rémi Letot) writes:

> If \Seen-marks are equivallent to read, I just subscribed a non critical
> group with old read and unread articles, did g, did M-g, entered it, all
> of these several times, and didn't see any modification in the marks.

*phew*  Thanks for testing.  :-)

> My server is an old Courrier, which I guess doesn't have QRESYNC ?

No, I don't think so.

So I believe the marking-everything-as-read problem experience by me
probably isn't real.  I hope.

-- 
(domestic pets only, the antidote for overdose, milk.)
  larsi@gnus.org * Lars Magne Ingebrigtsen

Re: nnimap danger! Do not use!

[Lars Magne Ingebrigtsen]

Steinar Bang <sb@dod.no> writes:

> I've sent the info for before and after for one group to you in email.

D'oh.  That was a pretty stupid logic error on my part.  In the QRESYNC
part, I didn't mark \Seen articles as read.  Fixed and pushed now.

-- 
(domestic pets only, the antidote for overdose, milk.)
  larsi@gnus.org * Lars Magne Ingebrigtsen

Re: nnimap danger! Do not use!

[Steinar Bang]

>>>>> Lars Magne Ingebrigtsen <larsi@gnus.org>:

> Steinar Bang <sb@dod.no> writes:
>> I've sent the info for before and after for one group to you in email.

> D'oh.  That was a pretty stupid logic error on my part.  In the QRESYNC
> part, I didn't mark \Seen articles as read.  Fixed and pushed now.

Something's still not working, but I'm not sure what and where things
are failing.


The conversation with you, last night, was read with a September 5
version of gnus on my desktop machine.  And the single new unread
article in the group I sent you the .newsrc.eld contents of, was read on
the desktop machine on a fresh git pull of this morning.

When reading on the netbook from a fresh git pull, the conversation
showed up as 8 unread articles.  And the one article I had read with a
fresh git pull'd gnus, also showed up as unread.

Is it a case of .newsrc.eld overriding the IMAP server?

Re: nnimap danger! Do not use!

[Lars Magne Ingebrigtsen]

Steinar Bang <sb@dod.no> writes:

> When reading on the netbook from a fresh git pull, the conversation
> showed up as 8 unread articles.  And the one article I had read with a
> fresh git pull'd gnus, also showed up as unread.

If you had `g'-d with a Gnus before the last checkin, then it won't
discover the changes that had happened before that change.  You need to
`M-g' to get a complete refresh, but after that it should then work as
advertised.

I hope.

-- 
(domestic pets only, the antidote for overdose, milk.)
  larsi@gnus.org * Lars Magne Ingebrigtsen

Re: nnimap danger! Do not use!

[Steinar Bang]

>>>>> Lars Magne Ingebrigtsen <larsi@gnus.org>:

> If you had `g'-d with a Gnus before the last checkin, then it won't
> discover the changes that had happened before that change.  You need
> to `M-g' to get a complete refresh, but after that it should then work
> as advertised.

Ok.  I started up a gnus installation that hadn't been running since
Friday, after first pulling from git and rebuilding.

It came up with unread articles on all nnimap folders that had received
email since Friday.  I did `M-g' on all of them and the unread count
went to 0.  Visiting the groups I could verify that messages had the
expected tick marks.

What does `M-g' do?  Update the .newsrc.eld information?  Update the
flags on the server?  If `M-g' is updating the .newsrc.eld, does that
mean that the .newsrc.eld information can override the the information
on the IMAP server?

Re: nnimap danger! Do not use!

[Lars Magne Ingebrigtsen]

Steinar Bang <sb@dod.no> writes:

> What does `M-g' do?  Update the .newsrc.eld information?

Yes.

nnimap groups work on the principle that the marks in the .newsrc.eld
should reflect what the server thinks of the messages.  However, getting
all the data on all the articles from the server is s-l-o-w, but this is
what `M-g' does.  `g' uses various other strategies for keeping the
IMAP->Gnus data flow in sync -- the latest of which is QRESYNC.  If
QRESYNC gets, er, out of sync, then Gnus won't know about the marks on
the server.

But since there are no bugs in nnimap now (ahem), that won't happen again.

-- 
(domestic pets only, the antidote for overdose, milk.)
  larsi@gnus.org * Lars Magne Ingebrigtsen

Re: nnimap danger! Do not use!

[Steinar Bang]

>>>>> Lars Magne Ingebrigtsen <larsi@gnus.org>:

> nnimap groups work on the principle that the marks in the .newsrc.eld
> should reflect what the server thinks of the messages.  However,
> getting all the data on all the articles from the server is s-l-o-w,
> but this is what `M-g' does.

So, if I had entered the group I would have seen the expected articles
ticked, this time around?

> `g' uses various other strategies for keeping the IMAP->Gnus data flow
> in sync -- the latest of which is QRESYNC.  If QRESYNC gets, er, out
> of sync, then Gnus won't know about the marks on the server.

Hum...?  I guess I will have to read up on what QRESYNC does.

> But since there are no bugs in nnimap now (ahem), that won't happen
> again.

No significant nnimap bugs, that a significant number of nnimap users,
would like fixed...?

Re: nnimap danger! Do not use!

[Lars Magne Ingebrigtsen]

Steinar Bang <sb@dod.no> writes:

> So, if I had entered the group I would have seen the expected articles
> ticked, this time around?

Yes.

>> `g' uses various other strategies for keeping the IMAP->Gnus data flow
>> in sync -- the latest of which is QRESYNC.  If QRESYNC gets, er, out
>> of sync, then Gnus won't know about the marks on the server.
>
> Hum...?  I guess I will have to read up on what QRESYNC does.

QRESYNC works by Gnus saying "give me all updates since sequence X" to
the IMAP server.  

>> But since there are no bugs in nnimap now (ahem), that won't happen
>> again.
>
> No significant nnimap bugs, that a significant number of nnimap users,
> would like fixed...?

No?  :-)

-- 
(domestic pets only, the antidote for overdose, milk.)
  larsi@gnus.org * Lars Magne Ingebrigtsen

Re: nnimap danger! Do not use!

[Rémi Letot]

Lars Magne Ingebrigtsen <larsi@gnus.org> writes:

> Steinar Bang <sb@dod.no> writes:
>
>
>>> But since there are no bugs in nnimap now (ahem), that won't happen
>>> again.
>>
>> No significant nnimap bugs, that a significant number of nnimap users,
>> would like fixed...?
>
> No?  :-)

Define "significant number" :-)

-- 
Rémi

Re: nnimap danger! Do not use!

[Steinar Bang]

>>>>> Lars Magne Ingebrigtsen <larsi@gnus.org>:

> Steinar Bang <sb@dod.no> writes:

>> No significant nnimap bugs, that a significant number of nnimap
>> users, would like fixed...?

> No?  :-)

That QRESYNC thing isn't working quite right for me yet, I think.  I
have to `M-g' to sync up, every time I switch to a different gnus.

Re: nnimap danger! Do not use!

[Lars Magne Ingebrigtsen]

Steinar Bang <sb@dod.no> writes:

> That QRESYNC thing isn't working quite right for me yet, I think.  I
> have to `M-g' to sync up, every time I switch to a different gnus.

Are all the Gnuses using different .newsrc.eld files?  

-- 
(domestic pets only, the antidote for overdose, milk.)
  larsi@gnus.org * Lars Magne Ingebrigtsen

Re: nnimap danger! Do not use!

[Steinar Bang]

>>>>> Lars Magne Ingebrigtsen <larsi@gnus.org>:

> Steinar Bang <sb@dod.no> writes:
>> That QRESYNC thing isn't working quite right for me yet, I think.  I
>> have to `M-g' to sync up, every time I switch to a different gnus.

> Are all the Gnuses using different .newsrc.eld files?  

Indeed they are.

They have an overlapping, but not identical, selection of NNTP servers,
the same single IMAP server, and selection of groups over the servers.

For the last couple of months I've been using Ted's gnus-sync.el to
synch the read marks for the NNTP groups.

But for the nnimap groups I would like the server to be the master.

Re: nnimap danger! Do not use!

[Lars Magne Ingebrigtsen]

Steinar Bang <sb@dod.no> writes:

>> Are all the Gnuses using different .newsrc.eld files?  
>
> Indeed they are.

Hm.  Then something is definitely not working in the QRESYNC case.

Could you try the following:

Start Gnus on the two machines.  Read and tick some articles on one of
them.  Then hit `g' in the other Gnus.  (And that other Gnus has to be a
fresh git one.)  Then look in the "*nnimap <server> ...*" buffer, and
post the output.  Also look in the "*imap log*" buffer for the commands
for that group, which should read something like

21:38:52 902 EXAMINE "MAILBOX.NAME" (QRESYNC (1 1))

or something.  

-- 
(domestic pets only, the antidote for overdose, milk.)
  larsi@gnus.org * Lars Magne Ingebrigtsen

Re: nnimap danger! Do not use!

[Steinar Bang]

>>>>> Lars Magne Ingebrigtsen <larsi@gnus.org>:

> Start Gnus on the two machines.  Read and tick some articles on one of
> them.  Then hit `g' in the other Gnus.  (And that other Gnus has to be a
> fresh git one.)  Then look in the "*nnimap <server> ...*" buffer, and
> post the output.  Also look in the "*imap log*" buffer for the commands
> for that group, which should read something like

> 21:38:52 902 EXAMINE "MAILBOX.NAME" (QRESYNC (1 1))

Well, eh, that turned into a "not repro" for me.  At least for a single
article read.

What I did:
 - Send a new message to sb@dod.no
 - Do repeated `g' on one gnus until I saw one unread, then enter the
   group and read the article and quit the group
 - Then do `g' on the other gnus (where demonical get hadn't
   discovered a new unread article yet), and the group showed up with no
   unread messages

Perhaps the QRESYNC things have eventually, er... synched up...?

Anyway.  Relevant line from *imap log*:
21:55:09 745 EXAMINE "dod" (QRESYNC (1221068358 419))

Here's the "*nnimap <server> ...*" buffer contents (with ^M suppressed):
* OK [CLOSED] Previous mailbox closed.
* FLAGS (%Answered %Flagged %Deleted %Seen %Draft)
* OK [PERMANENTFLAGS ()] Read-only mailbox.
* 17 EXISTS
* 5 RECENT
* OK [UNSEEN 2] First unseen.
* OK [UIDVALIDITY 1249481226] UIDs valid
* OK [UIDNEXT 385] Predicted next UID
* OK [HIGHESTMODSEQ 28] Highest
735 OK [READ-ONLY] Select completed.
* OK [CLOSED] Previous mailbox closed.
* FLAGS (%Answered %Flagged %Deleted %Seen %Draft gnus-forward gnus-dormant)
* OK [PERMANENTFLAGS ()] Read-only mailbox.
* 1892 EXISTS
* 0 RECENT
* OK [UIDVALIDITY 1221076465] UIDs valid
* OK [UIDNEXT 1896] Predicted next UID
* OK [HIGHESTMODSEQ 35] Highest
736 OK [READ-ONLY] Select completed.
* OK [CLOSED] Previous mailbox closed.
* FLAGS (%Answered %Flagged %Deleted %Seen %Draft)
* OK [PERMANENTFLAGS ()] Read-only mailbox.
* 2235 EXISTS
* 1 RECENT
* OK [UNSEEN 507] First unseen.
* OK [UIDVALIDITY 1221075897] UIDs valid
* OK [UIDNEXT 2236] Predicted next UID
* OK [HIGHESTMODSEQ 1] Highest
737 OK [READ-ONLY] Select completed.
* OK [CLOSED] Previous mailbox closed.
* FLAGS (%Answered %Flagged %Deleted %Seen %Draft gnus-forward gnus-dormant)
* OK [PERMANENTFLAGS ()] Read-only mailbox.
* 1274 EXISTS
* 0 RECENT
* OK [UIDVALIDITY 1221068129] UIDs valid
* OK [UIDNEXT 1283] Predicted next UID
* OK [HIGHESTMODSEQ 7] Highest
738 OK [READ-ONLY] Select completed.
* OK [CLOSED] Previous mailbox closed.
* FLAGS (%Answered %Flagged %Deleted %Seen %Draft gnus-forward)
* OK [PERMANENTFLAGS ()] Read-only mailbox.
* 1084 EXISTS
* 0 RECENT
* OK [UIDVALIDITY 1221076285] UIDs valid
* OK [UIDNEXT 1090] Predicted next UID
* OK [HIGHESTMODSEQ 1] Highest
739 OK [READ-ONLY] Select completed.
* OK [CLOSED] Previous mailbox closed.
* FLAGS (%Answered %Flagged %Deleted %Seen %Draft gnus-expire)
* OK [PERMANENTFLAGS ()] Read-only mailbox.
* 187 EXISTS
* 7 RECENT
* OK [UNSEEN 52] First unseen.
* OK [UIDVALIDITY 1221068004] UIDs valid
* OK [UIDNEXT 2107] Predicted next UID
* OK [HIGHESTMODSEQ 17] Highest
740 OK [READ-ONLY] Select completed.
* OK [CLOSED] Previous mailbox closed.
* FLAGS (%Answered %Flagged %Deleted %Seen %Draft)
* OK [PERMANENTFLAGS ()] Read-only mailbox.
* 190 EXISTS
* 0 RECENT
* OK [UIDVALIDITY 1284195970] UIDs valid
* OK [UIDNEXT 191] Predicted next UID
* OK [HIGHESTMODSEQ 3] Highest
741 OK [READ-ONLY] Select completed.
* OK [CLOSED] Previous mailbox closed.
* FLAGS (%Answered %Flagged %Deleted %Seen %Draft)
* OK [PERMANENTFLAGS ()] Read-only mailbox.
* 38 EXISTS
* 0 RECENT
* OK [UIDVALIDITY 1279388661] UIDs valid
* OK [UIDNEXT 48] Predicted next UID
* OK [HIGHESTMODSEQ 1] Highest
742 OK [READ-ONLY] Select completed.
* OK [CLOSED] Previous mailbox closed.
* FLAGS (%Answered %Flagged %Deleted %Seen %Draft)
* OK [PERMANENTFLAGS ()] Read-only mailbox.
* 57 EXISTS
* 0 RECENT
* OK [UIDVALIDITY 1279379802] UIDs valid
* OK [UIDNEXT 58] Predicted next UID
* OK [HIGHESTMODSEQ 1] Highest
743 OK [READ-ONLY] Select completed.
* OK [CLOSED] Previous mailbox closed.
* FLAGS (%Answered %Flagged %Deleted %Seen %Draft gnus-forward $Forwarded)
* OK [PERMANENTFLAGS ()] Read-only mailbox.
* 24161 EXISTS
* 0 RECENT
* OK [UIDVALIDITY 1221068973] UIDs valid
* OK [UIDNEXT 24165] Predicted next UID
* OK [HIGHESTMODSEQ 1] Highest
744 OK [READ-ONLY] Select completed.
* OK [CLOSED] Previous mailbox closed.
* FLAGS (%Answered %Flagged %Deleted %Seen %Draft gnus-forward $Forwarded)
* OK [PERMANENTFLAGS ()] Read-only mailbox.
* 593 EXISTS
* 0 RECENT
* OK [UIDVALIDITY 1221068358] UIDs valid
* OK [UIDNEXT 7276] Predicted next UID
* OK [HIGHESTMODSEQ 421] Highest
* 593 FETCH (MODSEQ (421) UID 7275 FLAGS (%Seen))
745 OK [READ-ONLY] Select completed.
* OK [CLOSED] Previous mailbox closed.
* FLAGS (%Answered %Flagged %Deleted %Seen %Draft)
* OK [PERMANENTFLAGS ()] Read-only mailbox.
* 587 EXISTS
* 0 RECENT
* OK [UIDVALIDITY 1221067842] UIDs valid
* OK [UIDNEXT 650] Predicted next UID
* OK [HIGHESTMODSEQ 36] Highest
746 OK [READ-ONLY] Select completed.

Re: nnimap danger! Do not use!

[Lars Magne Ingebrigtsen]

Steinar Bang <sb@dod.no> writes:

> Perhaps the QRESYNC things have eventually, er... synched up...?

Sounds like it.  :-)

> Anyway.  Relevant line from *imap log*:
> 21:55:09 745 EXAMINE "dod" (QRESYNC (1221068358 419))

[...]

> * OK [HIGHESTMODSEQ 421] Highest
> * 593 FETCH (MODSEQ (421) UID 7275 FLAGS (%Seen))

Looks just like it should, so I think this is probably working now.

-- 
(domestic pets only, the antidote for overdose, milk.)
  larsi@gnus.org * Lars Magne Ingebrigtsen