Make gnus-agent default to t?

Make gnus-agent default to t?

[Simon Josefsson]

I accidently changed the default of `nnimap-nov-is-evil' to t (fixed,
don't worry) because it is not needed when the Gnus Agent NOV cache is
used.  But now I discovered the Agent NOV cache isn't enabled unless
`gnus-agent' is non-nil.

What about making `gnus-agent' default to t?  It shouldn't do anything
bad, I think.  I've been setting it to t for ages without actually
using the agent for reading offline.  If something bad happens, maybe
that should be fixed instead.

The benefit would be that Gnus cached NOV headers for remote servers.
No more "nntp read: 4711kb" when entering groups.  I think this could
speed things up considerably.

Unfortunately you still need to add secondary servers to the Agent
manually though, but perhaps that could be automized as well?

Opinions?

Re: Make gnus-agent default to t?

[Kai Großjohann]

Simon Josefsson <jas@extundo.com> writes:

> What about making `gnus-agent' default to t?

Probably a good idea.  I have (gnus-agentize) in my .gnus, what's the
difference?

kai
-- 
Silence is foo!

Re: Make gnus-agent default to t?

[Simon Josefsson]

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

> Simon Josefsson <jas@extundo.com> writes:
>
>> What about making `gnus-agent' default to t?
>
> Probably a good idea.  I have (gnus-agentize) in my .gnus, what's the
> difference?

None. :-)

gnus-agent's value is t

Documentation:
Whether we want to use the Gnus agent or not.
Putting (gnus-agentize) in ~/.gnus is obsolete by (setq gnus-agent t).

You can customize this variable.

Defined in `gnus'.

Re: Make gnus-agent default to t?

[Simon Josefsson]

Simon Josefsson <jas@extundo.com> writes:

> Kai.Grossjohann@CS.Uni-Dortmund.DE (Kai writes:
>
>> Simon Josefsson <jas@extundo.com> writes:
>>
>>> What about making `gnus-agent' default to t?
>>
>> Probably a good idea.  I have (gnus-agentize) in my .gnus, what's the
>> difference?
>
> None. :-)
>
> gnus-agent's value is t

This is now the default.

Re: Make gnus-agent default to t?

[Florian Weimer]

Simon Josefsson <jas@extundo.com> writes:

>> gnus-agent's value is t
>
> This is now the default.

Doesn't this mean that mail fetching stops for people who get mail
from their local POP3 server or something like that?

Re: Make gnus-agent default to t?

[David S Goldberg]

>>>>> On Wed, 29 May 2002 21:29:06 +0200, Florian Weimer
>>>>> <fw@deneb.enyo.de> said: 

> Simon Josefsson <jas@extundo.com> writes:
>>> gnus-agent's value is t
>> 
>> This is now the default.

> Doesn't this mean that mail fetching stops for people who get mail
> from their local POP3 server or something like that?

Not in my experience.  In fact, when I'm truly unplugged, I have to
explicitly set nnml-get-new-mail to nil to avoid hanging while the
attempted POP3 connection times out.
-- 
Dave Goldberg
david.goldberg6@verizon.net

cc generated with both to-address and to-list set (was Re: Make gnus-agent default to t?)

[David S Goldberg]

Sorry, Kai, about the cc on that last message.  I thought I had my
group params set up so that F in my ding group would only ever send to
the mailing list by setting both to-list and to-address to
ding@gnus.org.  I don't recall a CC ever being generated to someone
else in groups with that setting before, though it's certainly
possible that I've not ever replied to a message with the headers set
as in Florian's message.  Do I misunderstand how to set this up
properly or did something change/break?

Thanks,
-- 
Dave Goldberg
david.goldberg6@verizon.net

Re: cc generated with both to-address and to-list set

[Paul Jarc]

David S Goldberg <david.goldberg6@verizon.net> wrote:
> Sorry, Kai, about the cc on that last message.  I thought I had my
> group params set up so that F in my ding group would only ever send to
> the mailing list by setting both to-list and to-address to
> ding@gnus.org.

Mail-Followup-To overrides to-address.  Kai doesn't set MFT, so
Simon's followup to Kai went to both Kai and ding.  Florian's followup
to Simon thus also went to both Kai and ding (but not Simon, because
Simon sets Mail-Copies-To: nobody).  Florian does set MFT, so MFT
included Kai and ding.  Thus your followup to Florian included Kai and
ding.

Everyone should set MFT.  Then this won't happen, and to-address will
be unnecessary.
(setq message-subscribed-address-functions '(gnus-find-subscribed-addresses))
(add-to-list gnus-parameters '(("ding" (subscribed . t))))


paul

Re: cc generated with both to-address and to-list set

[Benjamin Rutt]

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

> Everyone should set MFT.  Then this won't happen, and to-address will
> be unnecessary.

I'd argue that to-address should be avoided regardless, since it
prevents non-subscribers from receiving followups to their own
postings.  Setting to-address is like asserting "I know that all the
people reading the list are subscribed".  I've never read a mailing
list where I could prove that assertion.

I set MFT (on mailing lists that is...I'm reading ding via nntp from
news.gnus.org) and agree that it solves the problem but I don't expect
it to heal the world overnight.  So until the rest of the world
catches up, I (setq nnmail-treat-duplicates 'delete).

I guess setting nnmail-treat-duplicates to 'delete won't save me from
getting extra potential followups to this message, since I'm reading
via nntp!  Maybe it's time to switch to receiving the mailing list
just for that reason.  I guess another reason would be the ability to
usefully type 'A T' in Summary (i.e. without waiting forever for it to
return).
-- 
Benjamin

Re: cc generated with both to-address and to-list set

[Per Abrahamsen]

Benjamin Rutt <rutt+news@cis.ohio-state.edu> writes:

> postings.  Setting to-address is like asserting "I know that all the
> people reading the list are subscribed".  I've never read a mailing
> list where I could prove that assertion.

Most mailing list software has a "disallow non-subscribers posting"
option, which I guess is getting increasingly common in order to
prevent spam.

The USEFOR list used to work like that, until someone pointed out it
was against the IETF rules.

Re: cc generated with both to-address and to-list set

[Paul Jarc]

Per Abrahamsen <abraham@dina.kvl.dk> wrote:
> Most mailing list software has a "disallow non-subscribers posting"
> option, which I guess is getting increasingly common in order to
> prevent spam.

The good ones also allow people to register for posting without
subscribing.  Anyway, even if someone is subscribed, it's possible
they might want an extra copy for some reason.  Only MFT can tell us
whether that is the case.


paul

Re: Make gnus-agent default to t?

[Simon Josefsson]

Florian Weimer <fw@deneb.enyo.de> writes:

> Simon Josefsson <jas@extundo.com> writes:
>
>>> gnus-agent's value is t
>>
>> This is now the default.
>
> Doesn't this mean that mail fetching stops for people who get mail
> from their local POP3 server or something like that?

That was not the intention, does it happen?  The Agent is in the
Plugged state by default, in which case everything is supposed to work
like before, I think.

The goal was to 1) make Gnus use the Agent for NOV and article caching
like many other MUAs and 2) make the Agent more integrated with Gnus
in general (to start using the Agent, all that should be required is
to fetch a group using J u).

(Shouldn't the To/CC of this mail contain both Kai and ding on the To:
line?  Why did it put Kai on the To: and ding on Cc?)

Re: Make gnus-agent default to t?

[Paul Jarc]

Simon Josefsson <jas@extundo.com> wrote:
> (Shouldn't the To/CC of this mail contain both Kai and ding on the To:
> line?  Why did it put Kai on the To: and ding on Cc?)

The recipient list (as determined by Florian's MFT) contains Kai and
ding, in that order.  message-get-reply-headers puts the first
recipient in To and the rest in Cc.


paul

Re: Make gnus-agent default to t?

[Paul Jarc]

Simon Josefsson <jas@extundo.com> wrote:
>> gnus-agent's value is t
>
> This is now the default.

nnmaildir users may want to set this to nil for a while; I'm still
working on making article numbers persistent, so header caching could
break right now.


paul

Re: Make gnus-agent default to t?

[Simon Josefsson]

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

> Simon Josefsson <jas@extundo.com> wrote:
>>> gnus-agent's value is t
>>
>> This is now the default.
>
> nnmaildir users may want to set this to nil for a while; I'm still
> working on making article numbers persistent, so header caching could
> break right now.

I forgot about that.  Hm.  Perhaps extending `gnus-agent-cache' so it
can exclude certain groups would be useful?  Then it could default to
cache everything for all groups except nnmaildir groups (until
nnmaildir gets persistent article numbers).  What do you think?

Re: Make gnus-agent default to t?

[Paul Jarc]

Simon Josefsson <jas@extundo.com> wrote:
> Perhaps extending `gnus-agent-cache' so it can exclude certain
> groups would be useful?  Then it could default to cache everything
> for all groups except nnmaildir groups (until nnmaildir gets
> persistent article numbers).  What do you think?

I doubt it would be worth the effort.  nnmaildir should be fixed RSN.


paul

Re: Make gnus-agent default to t?

[Simon Josefsson]

Simon Josefsson <jas@extundo.com> writes:

> prj@po.cwru.edu (Paul Jarc) writes:
>
>> Simon Josefsson <jas@extundo.com> wrote:
>>>> gnus-agent's value is t
>>>
>>> This is now the default.
>>
>> nnmaildir users may want to set this to nil for a while; I'm still
>> working on making article numbers persistent, so header caching could
>> break right now.
>
> I forgot about that.  Hm.  Perhaps extending `gnus-agent-cache' so it
> can exclude certain groups would be useful?  Then it could default to
> cache everything for all groups except nnmaildir groups (until
> nnmaildir gets persistent article numbers).  What do you think?

Never mind, there's no need for panic, or setting `gnus-agent' to nil:

Agent NOV caching doesn't make sense for local servers at all, does
it?

I think nnmaildir users would only have problems if they had agentized
the nnmaildir server.  Since that didn't work in the past (the agent
really require persistant article numbers), there is no change.

Don't add nnmaildir servers to the agent.

Try running nnmaildir with `gnus-agent' at its default value.  There
should be no ~/News/agent/nnmaildir hierarchy.  If there is, there
will be problems.  Otherwise, you're safe.

OTOH users must enroll their remote secondary nntp and nnimap servers
with the agent to gain the NOV caching feature.  This should probably
happen by default, but only for nntp and nnimap?

Hm.  Agent NOV caching would obsolete nnfolder NOVs, wouldn't it?

Re: Make gnus-agent default to t?

[Kai Großjohann]

Simon Josefsson <jas@extundo.com> writes:

> OTOH users must enroll their remote secondary nntp and nnimap servers
> with the agent to gain the NOV caching feature.  This should probably
> happen by default, but only for nntp and nnimap?

Is the native server automatically added to the agent?

kai
-- 
Silence is foo!

Re: Make gnus-agent default to t?

[Simon Josefsson]

Kai.Grossjohann@CS.Uni-Dortmund.DE (Kai Großjohann) writes:

> Simon Josefsson <jas@extundo.com> writes:
>
>> OTOH users must enroll their remote secondary nntp and nnimap servers
>> with the agent to gain the NOV caching feature.  This should probably
>> happen by default, but only for nntp and nnimap?
>
> Is the native server automatically added to the agent?

Yes.  So if you have a nnmaildir as the primary server, you should
de-enroll (deroll?) it from the agent until Paul modifies nnmaildir.

Re: Make gnus-agent default to t?

[Christoph Rohland]

Hi Simon,

On Sat, 25 May 2002, Simon Josefsson wrote:
> What about making `gnus-agent' default to t?  It shouldn't do
> anything bad, I think.  I've been setting it to t for ages without
> actually using the agent for reading offline.  If something bad
> happens, maybe that should be fixed instead.

I reported the following bad behaviour of the current agent which
prevents me from using oort gnus:

If you 'B Del' an article it shows up later in the summary. Wes
Hardaker did proposed to regularaly run gnus-agent-expire which is
IMNSHO mere bandaid. I do consider this a real bug which is AFAIK not
fixed and should prevent you from setting gnus-agent to t per default.

Greetings
		Christoph

Re: Make gnus-agent default to t?

[Simon Josefsson]

Christoph Rohland <cr@sap.com> writes:

> Hi Simon,
>
> On Sat, 25 May 2002, Simon Josefsson wrote:
>> What about making `gnus-agent' default to t?  It shouldn't do
>> anything bad, I think.  I've been setting it to t for ages without
>> actually using the agent for reading offline.  If something bad
>> happens, maybe that should be fixed instead.
>
> I reported the following bad behaviour of the current agent which
> prevents me from using oort gnus:
>
> If you 'B Del' an article it shows up later in the summary. Wes
> Hardaker did proposed to regularaly run gnus-agent-expire which is
> IMNSHO mere bandaid. I do consider this a real bug which is AFAIK not
> fixed and should prevent you from setting gnus-agent to t per default.

Yes, this has bugged me for a while as well. (To workaround the
problem I have been removing the .overview file in the Agent directory
for the group, as it is rebuilt if missing).

I tried to fix it in Oort, but if someone who understands
`gnus-agent-expire' would take a look at it to see if it is correct,
it would be good.  If someone has problems with Agent Expiry or B DEL
now, please speak up.

I'll fix the doc later, unless someone discoveres problems that cannot
easily be fixed and we'll have to revert to the old behaviour.