gnus and POP3

gnus and POP3

[Stefan Waldherr]

Hi,

is there any way to tell gnus to use POP3 _and_ leave the mail on the server
(without using any non-gnus, non-emacs stuff)? gnus should retrieve only _new_
mail from the server and should leave the mail where it is.

Thanks,
Stefan.
-- 
Stefan Waldherr                   fax +49 431 8058 136
                               e-Mail stefan@waldherr.org
                                  www http://www.waldherr.org/

Re: gnus and POP3

[Wes Hardaker]

>>>>> On 16 Jan 1999 21:38:55 +0100, Stefan Waldherr <stefan@waldherr.org> said:

Stefan> is there any way to tell gnus to use POP3 _and_ leave the mail
Stefan> on the server (without using any non-gnus, non-emacs stuff)?
Stefan> gnus should retrieve only _new_ mail from the server and
Stefan> should leave the mail where it is.

I hacked pop3.el a long time ago to do this.  I submitted the patch to 
Lars and the pop3 author.  They told me I did it all wrong (not
surprising...  learning the layout of everything is a very steep curve
when it comes to gnus).  Anyway, the proper way to do it is to create
a nnmail-delete-incoming variable.  Not like the pop3-delete-incoming
variable I use now.  It's a long needed feature.  I can send you the
patch if you like.  I love CVS.  I have about 3 patches for features I 
wrote that have never been properly integrated into gnus due to
they're usually the "wrong way to do it" (and they are...  But I don't 
have the 4 hours of learning time available to do them correctly)...

(sigh...  one of these days I'll no longer need sleep).
-- 
"Ninjas aren't dangerous.  They're more afraid of you than you are of them."

Re: gnus and POP3

[Wes Hardaker]

>>>>> On 17 Jan 1999 19:27:35 -0500, Stainless Steel Rat <ratinox@peorth.gweep.net> said:

WH> I hacked pop3.el a long time ago to do this.  I submitted the
WH> patch to Lars and the pop3 author.  They told me I did it all
WH> wrong (not surprising...

Rat> For two reasons.

Rat> One, the more important reason, is that adding features to
Rat> pop3-movemail is a Bad Thing.

Agreed.

Rat> pop3-movemail is a reference implementation demonstrating how to
Rat> use the POP library.

Yep.  And its in use by a whole huge ton of people (I use it over the
C version since it has better feedback and apop support).

Rat> If you want a more featureful movemail function, please, write
Rat> one.

If I get the time, I promise to.  Sadly, I'm on a year and a half
project that is blowing all my spare time working towards another PD
project so my gnus hacking came to a halt a while back.

Rat> Two, the more important reason, 'leave mail on server' is not a
Rat> feature, but a misuse and abuse of the POP3 protocol that puts an
Rat> unnecessary strain on the server: every time you open a POP
Rat> connection, the POP server program scans the entire mailbox.

Yep.

Rat> Doing this with large mail files can put a serious strain on the
Rat> server's memory and disk I/O.  I have seen a P90 with 32Mb RAM
Rat> and 96Mb swap, running Linux, utterly throttled in this fashion.

Fortunately, I'm using a much better machine with a lot more memory
and only 30 users on it (total...  3-4 at a time), and only a couple
make use of the pop server.  I won't overload it.

Rat> Use IMAP if you are going to do this; that is one of the main
Rat> reasons it was invented.

Yep.  Agreed.  And the nnimap implementation is coming along.  I've
been playing with it for a while.  However, it does not support the
agent yet, which I *must* have since I pay per-minute charges for my
isdn line from 8-5.  I've looked into the changes required, but they
are too vast to handle myself at the moment.  The other problem is
that recently the nnimap backend has started causing my xemacs to run
away consuming all memory on my machine till I kill xemacs.  It's
apparently a known problem.  Hopefully the re-write that is happening
of the nnimap backend will help this situation.  In the mean time, I'm 
stuck.  I have to have leave-on-server support when I'm at work (which 
is 2 days a week) and download to disk 5 days a week (when at home).
This is the only feasible solution until nnimap becomes stable enough
to use.  I suspect we're still a few months away at least from that...

So, in the mean time, I'll be sticking with my leave-mail-on-server
hack...  Sadly.  I'd love not to use it...
-- 
"Ninjas aren't dangerous.  They're more afraid of you than you are of them."

Re: gnus and POP3

[Hrvoje Niksic]

Wes Hardaker <wjhardaker@ucdavis.edu> writes:

> The other problem is that recently the nnimap backend has started
> causing my xemacs to run away consuming all memory on my machine
> till I kill xemacs.

Is this an XEmacs bug?  I haven't seen it reported.

Re: gnus and POP3

[Jack Vinson]

>>>>> "SW" == Stefan Waldherr <stefan@waldherr.org> writes:

SW> is there any way to tell gnus to use POP3 _and_ leave the mail on the
SW> server (without using any non-gnus, non-emacs stuff)? gnus should
SW> retrieve only _new_ mail from the server and should leave the mail
SW> where it is.

As someone (Richard?) mentioned, Franklin Lee <flee@lehman.com> has written
epop3mail (extension to pop3.el) that permits exactly this functionality.
He has also built in a biff function,

His purpose in writing the code was to provide more flexibility than the
default movemail functionality.  The example he gives is someone who is on
a temporary machine who does not want to lose the mail that normally goes
to his main machine.

I have been using this for about six months with little difficulty.
Actually, we've discovered a few little problems:  It isn't treating
embedded messages properly, and it doesn't permit type-ahead for some
reason.  

Here is a link to this package:
http://www.cs.washington.edu/homes/voelker/ntemacs/contrib/ep3m-096.zip

-- 
Jack Vinson <jvinson@chevax.ecs.umass.edu>    http://www.cis.upenn.edu/~vinson/
Zippy: I'm shaving!!  I'M SHAVING!!

Re: gnus and POP3

[Wes Hardaker]

>>>>> On 18 Jan 1999 19:00:28 -0500, Stainless Steel Rat <ratinox@peorth.gweep.net> said:

WH> Yep.  And its in use by a whole huge ton of people [...]

Rat> I am amazed at this, I truely am.

Just goes to show you haw badly alternatives are needed ;-)

-- 
"Ninjas aren't dangerous.  They're more afraid of you than you are of them."

Re: gnus and POP3

[Wes Hardaker]

>>>>> On 19 Jan 1999 01:07:33 +0100, Hrvoje Niksic <hniksic@srce.hr> said:

>> The other problem is that recently the nnimap backend has started
>> causing my xemacs to run away consuming all memory on my machine
>> till I kill xemacs.

Hrvoje> Is this an XEmacs bug?  I haven't seen it reported.

I don't think so.  I think its an nnimap bug.  Ok, the memory
utilization might be.

I think it was happening with 20.4 as well as what I'm using now
(21.2.8), but I don't remember exactly (sorry).

I haven't reported it since I haven't had the time to trace it down
better yet.

-- 
"Ninjas aren't dangerous.  They're more afraid of you than you are of them."

Re: gnus and POP3

[Kai.Grossjohann]

Wes Hardaker <wjhardaker@ucdavis.edu> writes:

  > [...] However, it does not support the agent yet, which I *must*
  > have since I pay per-minute charges for my isdn line from
  > 8-5. [...]

In the meantime, you could remote login to work, tell Emacs to suck
mail there, then use rsync (or whatever) to synchronize ~/Mail at work
with ~/Mail at home (plus ~/News, ~/.newsrc* and ~/.nnmail-cache, of
course).  Would that be a q&d solution?

kai
-- 
Abort this operation?   [Abort]  [Cancel]

Re: gnus and POP3

[Wes Hardaker]

>>>>> On 20 Jan 1999 22:37:48 +0100, Kai.Grossjohann@CS.Uni-Dortmund.DE said:

>> [...] However, it does not support the agent yet, which I *must*
>> have since I pay per-minute charges for my isdn line from
>> 8-5. [...]

Kai> In the meantime, you could remote login to work, tell Emacs to
Kai> suck mail there, then use rsync (or whatever) to synchronize
Kai> ~/Mail at work with ~/Mail at home (plus ~/News, ~/.newsrc* and
Kai> ~/.nnmail-cache, of course).  Would that be a q&d solution?

Ick.  Yeah, I suppose its doable....  Though at times I have gnus
running in both places (I usually stay logged in at work, and it does
expiraries all night ;-).

Um, actually, no it won't work without a whole bunch of pain.  I
forgot...  I use different nntp servers, so synching the .newsrc files 
would be a bit difficult...

-- 
"Ninjas aren't dangerous.  They're more afraid of you than you are of them."

Re: gnus and POP3

[Kai.Grossjohann]

Wes Hardaker <wjhardaker@ucdavis.edu> writes:

  > Um, actually, no it won't work without a whole bunch of pain.  I
  > forgot...  I use different nntp servers, so synching the .newsrc
  > files would be a bit difficult...

You can't have two Gnusae running at the same time.  But you can
twiddle gnus-select-method and take advantage of the fact that Gnus
will use the files ~/.newsrc-foo* if they exist (and
gnus-select-method points to the foo machine).

This would mean that you need to exit Gnus then start it with the
other setting to switch from work to home mail/news.

kai
-- 
Abort this operation?   [Abort]  [Cancel]

Re: gnus and POP3

[Hrvoje Niksic]

Kai.Grossjohann@CS.Uni-Dortmund.DE writes:

> You can't have two Gnusae running at the same time.

Sure you can; `M-x gnus-slave' is your friend.

Re: gnus and POP3

[Steinar Bang]

>>>>> Stainless Steel Rat <ratinox@peorth.gweep.net>:

> "RC" == Richard Coleman <coleman@math.gatech.edu> writes:

RC> While I agree that IMAP is better designed for the given situation, the
RC> fact is that most people do not have a choice on whether to use POP or
RC> IMAP.  They must use what is provided for them.

> BS.  Most people are paying for their network services.  The onus is
> on the provider to meet the needs of its paying customers.  If the
> paying customers want IMAP, even if they do not know it by name, the
> provider should provide it.

Agreed!  Users should be trained to ask their ISPs forcefully for
IMAP.  

One problem with suggesting this earlier, have been a lack of good
IMAP clients.  However the recent versions of Netscape haven't been to 
shabby (4.04 was *bad* as I recall, but 4.5 behaved surprisingly well
the last time I tried it).

And of course, we will soon have nnimap... (thanx Simon!) :-) 

Re: gnus and POP3

[Richard Coleman]

> WH> I hacked pop3.el a long time ago to do this.  I submitted the patch to
> WH> Lars and the pop3 author.  They told me I did it all wrong (not
> WH> surprising...
> 
> For two reasons.
> 
> One, the more important reason, is that adding features to pop3-movemail is
> a Bad Thing.  pop3-movemail is a reference implementation demonstrating how
> to use the POP library.  If you want a more featureful movemail function,
> please, write one.  Or look at what Franklin Lee did in this regard with
> epop3mail.

No argument here.

> Two, the more important reason, 'leave mail on server' is not a feature,
> but a misuse and abuse of the POP3 protocol that puts an unnecessary strain
> on the server: every time you open a POP connection, the POP server program
> scans the entire mailbox.  Doing this with large mail files can put a
> serious strain on the server's memory and disk I/O.  I have seen a P90 with
> 32Mb RAM and 96Mb swap, running Linux, utterly throttled in this fashion.
> Use IMAP if you are going to do this; that is one of the main reasons it
> was invented.

Regardless of whether this was part of the original design or not, the
fact that this is *repeatedly* asked for, is a very important fact.
People always use (and abuse) software for purposes it wasn't
originally designed for.

While I agree that IMAP is better designed for the given situation, the
fact is that most people do not have a choice on whether to use POP or
IMAP.  They must use what is provided for them.

Also, this begs the question of whether a software designer is supposed
to provide features that customers want, or features that the designer
things the customer should have.

I don't want anyone to think I'm advocating the Microsoft mentality of
pampering users with fluff.  But I do think the wishes of the users
should strongly considered.

Now, since we are talking about free software, there are no real
"customers" per say.  I think if someone wants this functionality bad
enough he will write it.  But the canned response "just use IMAP" seems
pretty hollow.

--
Richard Coleman
coleman@math.gatech.edu