Kevin's gnus-open-server changes

Kevin's gnus-open-server changes

[Kai Großjohann]

Lars dropped the changes on the floor, which might have been a good
idea such a short time before the release.

But I understand that there was a reason for Kevin's changes.  So
what was the reason?  I wonder if there is a need for putting the
changes back (after 5.10 has been released during the 5.10 bugfix cycle).
-- 
file-error; Data: (Opening input file no such file or directory ~/.signature)

Re: Kevin's gnus-open-server changes

[Lars Magne Ingebrigtsen]

kai.grossjohann@gmx.net (Kai Großjohann) writes:

> But I understand that there was a reason for Kevin's changes.  So
> what was the reason? 

Gnus' unpluggedness is quite different from what you get if you click
on the "go offline" button in Free Agent or other offline
newsreaders.  There "go offline" means "I now no longer have a
network connections, so just do everything locally".  In Gnus it
means "inhibit the servers that are covered by the Agent from opening
network connections".

Kevin's changes made Gnus behave more like one might expect, but it's
something that can be hashed out in No Gnus.

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

Re: Kevin's gnus-open-server changes

[Xavier Maillard]

[-- Attachment #1: Type: text/plain, Size: 873 bytes --]

On  1 mai 2003, Lars Magne Ingebrigtsen wrote:

>  kai.grossjohann@gmx.net (Kai Großjohann) writes:
>  
> >  But I understand that there was a reason for Kevin's changes.  So
> >  what was the reason? 
>  
>  Gnus' unpluggedness is quite different from what you get if you click
>  on the "go offline" button in Free Agent or other offline
>  newsreaders.  There "go offline" means "I now no longer have a
>  network connections, so just do everything locally".  In Gnus it
>  means "inhibit the servers that are covered by the Agent from opening
>  network connections".
>  
>  Kevin's changes made Gnus behave more like one might expect, but it's
>  something that can be hashed out in No Gnus.

I agree with that.

zeDek
-- 
http://www.gnusfr.org -- French Gnus user site

Anti-war disclaimer:
	"Bombing for peace is like fucking for virginity"

[-- Attachment #2: Type: application/pgp-signature, Size: 188 bytes --]

Re: Kevin's gnus-open-server changes

[Kevin Greiner]

Xavier Maillard <zedek@gnu-rox.org> writes:

> On  1 mai 2003, Lars Magne Ingebrigtsen wrote:
>
>>  kai.grossjohann@gmx.net (Kai Großjohann) writes:
>>  
>> >  But I understand that there was a reason for Kevin's changes.  So
>> >  what was the reason? 

The thread <86el3lqrel.fsf@sophia.inria.fr> in gnus.ding.

I verified Max's observation by executing (gnus-unplugged) while I had
a physical connection to the net.  The result was that my unagentized
servers still opened their connections.  That appeared to be an error
given my reading of the manual (see below).  Since I had modified
gnus-open-server just a couple of months ago, I concluded that I had
introduced a bug.
  
>>  Gnus' unpluggedness is quite different from what you get if you click
>>  on the "go offline" button in Free Agent or other offline
>>  newsreaders.  There "go offline" means "I now no longer have a
>>  network connections, so just do everything locally".  In Gnus it
>>  means "inhibit the servers that are covered by the Agent from opening
>>  network connections".
>>  
>>  Kevin's changes made Gnus behave more like one might expect, but it's
>>  something that can be hashed out in No Gnus.
>
> I agree with that.

I too.

Now then, the first paragraph of Agent Basics in the manual reads:

  First, let's get some terminology out of the way.

  The Gnus Agent is said to be "unplugged" when you have severed the
  connection to the net (and notified the Agent that this is the case).
  When the connection to the net is up again (and Gnus knows this), the
  Agent is "plugged".

I read this description to mean that a server's state when unplugged
must be either denied or offline as those are the only possible states
when dealing with a connection-based server on an isolated machine.

What I didn't consider were

Should the plugged/unplugged status have any effect on a
connection-less server?  For example, shouldn't the nndoc backend always
have a status of ok?

Should the plugged/unplugged status have any effect on an unagentized
server?  If I'm physically connected, but unplugged, shouldn't the
status of an unagentized nntp server be ok (if the plugged status is
ignored) or denied (if the plugged status applies equally to all
servers)?

From yet another thread, if I have an unagentized (or even agentized)
mail server, should the agent ALWAYS store my messages in the drafts
folder?  It seems that some users have a local MTA so they'd like the
option of sending mail even when unplugged.

I hadn't thought to write down my definition of unplugged but it would
probably have been:

  Gnus is said to be "unplugged" when you have informed Gnus that it
  should close existing, and refuse to open new, connections to the
  net.  Connection-less servers are not affected by the "unplugged"
  status.

  Individual connection-based servers may be configured to ignore the
  "unplugged" status.  These servers will, of course, fail if Gnus is
  both "unplugged" and physically isolated from the net.

  Gnus is said to be "plugged" once it is again permitted to open
  connections.

The middle part hasn't been written as I was honestly trying to
introduce a bug fix rather than an entire new feature.  If it did get
written, I'd probably model it on the plugged keyword already existing
in the mail sources list.

So, those are my jumbled thoughts, I'd like to hear what others have
to think.

Kevin

Re: Kevin's gnus-open-server changes

[Kai Großjohann]

Kevin Greiner <kgreiner@xpediantsolutions.com> writes:

> I verified Max's observation by executing (gnus-unplugged) while I had
> a physical connection to the net.  The result was that my unagentized
> servers still opened their connections.  That appeared to be an error
> given my reading of the manual (see below).  Since I had modified
> gnus-open-server just a couple of months ago, I concluded that I had
> introduced a bug.

The manual only says what the agent does.  But the agent isn't
involved at all for those groups that are un-agentized.

The manual is hard to understand and possibly misleading, but it's
not really wrong.

I like the current behavior, so my suggestion is to fix the
documentation so that it explains better what is going on.

Maybe it would also be nice to offer an easy way of agentizing all
servers where that makes sense (ie, those which talk over the
network).  Something like a gnus-agent-add-all-servers command in the
*Servers* buffer, perhaps.

-- 
file-error; Data: (Opening input file no such file or directory ~/.signature)

Re: Kevin's gnus-open-server changes

[Lars Magne Ingebrigtsen]

kai.grossjohann@gmx.net (Kai Großjohann) writes:

> The manual is hard to understand and possibly misleading, but it's
> not really wrong.

:-)

Since the Gnus Agent doesn't do this the same way that other offline
readers does this, the manual should discuss that.  Patches, anyone?

> Maybe it would also be nice to offer an easy way of agentizing all
> servers where that makes sense (ie, those which talk over the
> network).  Something like a gnus-agent-add-all-servers command in the
> *Servers* buffer, perhaps.

Or just perhaps a command/toggle that would make `gnus-open-server'
work the way Kevin's changes made it work -- deny all servers from
trying to establish a network connection (whatever that may be) when
you flip the toggle.

(In general, this is an impossible task.  nnml may access stuff over
nfs/afs, but even if we ignore that, any of the mail backends may
include ange-ftp/tramp paths, which means that they really will do
network traffic even though you shouldn't think so.

But I think we may just ignore that complication.)

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

Re: Kevin's gnus-open-server changes

[Lars Magne Ingebrigtsen]

Kevin Greiner <kgreiner@xpediantsolutions.com> writes:

> Should the plugged/unplugged status have any effect on a
> connection-less server?  For example, shouldn't the nndoc backend always
> have a status of ok?

Well, nndoc might read something over ftp...

> I hadn't thought to write down my definition of unplugged but it would
> probably have been:
>
>   Gnus is said to be "unplugged" when you have informed Gnus that it
>   should close existing, and refuse to open new, connections to the
>   net.  Connection-less servers are not affected by the "unplugged"
>   status.
>
>   Individual connection-based servers may be configured to ignore the
>   "unplugged" status.  These servers will, of course, fail if Gnus is
>   both "unplugged" and physically isolated from the net.
>
>   Gnus is said to be "plugged" once it is again permitted to open
>   connections.
>
> The middle part hasn't been written as I was honestly trying to
> introduce a bug fix rather than an entire new feature.  If it did get
> written, I'd probably model it on the plugged keyword already existing
> in the mail sources list.
>
> So, those are my jumbled thoughts, I'd like to hear what others have
> to think.

It makes sense, and should be provided as an option for users that
want "total" unpluggedness without having the Agent cover all the
servers.  (In No Gnus.  :-)

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

Re: Kevin's gnus-open-server changes

[Marco Lonsing]

Kevin Greiner <kgreiner@xpediantsolutions.com> writes:

> Xavier Maillard <zedek@gnu-rox.org> writes:
>
>> On  1 mai 2003, Lars Magne Ingebrigtsen wrote:
>>
>>>  kai.grossjohann@gmx.net (Kai Großjohann) writes:
>>>  
>>> >  But I understand that there was a reason for Kevin's changes.  So
>>> >  what was the reason? 
>
> The thread <86el3lqrel.fsf@sophia.inria.fr> in gnus.ding.
>
> I verified Max's observation by executing (gnus-unplugged) while I had
> a physical connection to the net.  The result was that my unagentized
> servers still opened their connections.  That appeared to be an error
> given my reading of the manual (see below).  Since I had modified
> gnus-open-server just a couple of months ago, I concluded that I had
> introduced a bug.
>   
[...]

I had the same problem in the last weeks. I'm using two different nntp
servers which are both covered by the agent, while my nnml servers are not
agentized. If I start gnus-unplugged at home I always have to stop windows
starting an internet connection.


ML

Re: Kevin's gnus-open-server changes

[Harry Putnam]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain; charset=us-ascii, Size: 938 bytes --]

kai.grossjohann@gmx.net (Kai Großjohann) writes:

> I like the current behavior, so my suggestion is to fix the
> documentation so that it explains better what is going on.

I agree with Kai.  The behavior Kevin described has been there since
quassia days.  It makes sense that if a group is not agentized then
gnus should connect as usual.

Lars Magne Ingebrigtsen <larsi@gnus.org> writes:
> :-)
>
> Since the Gnus Agent doesn't do this the same way that other offline
> readers does this, the manual should discuss that.  Patches, anyone?
>
>> Maybe it would also be nice to offer an easy way of agentizing all
>> servers where that makes sense (ie, those which talk over the
>> network).  Something like a gnus-agent-add-all-servers command in the
>> *Servers* buffer, perhaps.

Hard to see why this is necessary.  A few words in the manual
pointing out that ONLY agentized groups plug and unplug should be
sufficient shouldn't it?

Re: Kevin's gnus-open-server changes

[Kai Großjohann]

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

> kai.grossjohann@gmx.net (Kai Großjohann) writes:
>
>> The manual is hard to understand and possibly misleading, but it's
>> not really wrong.
>
> :-)
>
> Since the Gnus Agent doesn't do this the same way that other offline
> readers does this, the manual should discuss that.  Patches, anyone?

Committed.

>> Maybe it would also be nice to offer an easy way of agentizing all
>> servers where that makes sense (ie, those which talk over the
>> network).  Something like a gnus-agent-add-all-servers command in the
>> *Servers* buffer, perhaps.
>
> Or just perhaps a command/toggle that would make `gnus-open-server'
> work the way Kevin's changes made it work -- deny all servers from
> trying to establish a network connection (whatever that may be) when
> you flip the toggle.

Whee.  Is this really useful?

-- 
file-error; Data: (Opening input file no such file or directory ~/.signature)

Re: Kevin's gnus-open-server changes

[Lars Magne Ingebrigtsen]

kai.grossjohann@gmx.net (Kai Großjohann) writes:

>> Or just perhaps a command/toggle that would make `gnus-open-server'
>> work the way Kevin's changes made it work -- deny all servers from
>> trying to establish a network connection (whatever that may be) when
>> you flip the toggle.
>
> Whee.  Is this really useful?

Say you have lots of different groups from different on your laptop,
but only agentize, say, nnimap, because those are the only ones you
want to work with while you're on the train.  So you work as normal
when connected to the net, hit a toggle, unplug the laptop and go on
answering your mail while on the train.  You return home, flip the
toggle, and carry on.

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

Re: Kevin's gnus-open-server changes

[Kai Großjohann]

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

> Say you have lots of different groups from different on your laptop,
> but only agentize, say, nnimap, because those are the only ones you
> want to work with while you're on the train.  So you work as normal
> when connected to the net, hit a toggle, unplug the laptop and go on
> answering your mail while on the train.  You return home, flip the
> toggle, and carry on.

Ah, I see.  I think that currently, Gnus will timeout on accessing
the net and ask the user whether they want to switch the server to
offline mode.  Is this correct?  Or is offline mode only used for
agentized servers?

-- 
file-error; Data: (Opening input file no such file or directory ~/.signature)

Re: Kevin's gnus-open-server changes

[Kevin Greiner]

kai.grossjohann@gmx.net (Kai Großjohann) writes:

> Lars Magne Ingebrigtsen <larsi@gnus.org> writes:
>
>> Say you have lots of different groups from different on your laptop,
>> but only agentize, say, nnimap, because those are the only ones you
>> want to work with while you're on the train.  So you work as normal
>> when connected to the net, hit a toggle, unplug the laptop and go on
>> answering your mail while on the train.  You return home, flip the
>> toggle, and carry on.
>
> Ah, I see.  I think that currently, Gnus will timeout on accessing
> the net and ask the user whether they want to switch the server to
> offline mode.  Is this correct?

That's one possible behavior.

(defcustom gnus-server-unopen-status nil
  "The default status if the server is not able to open.
If the server is covered by Gnus agent, the possible values are
`denied', set the server denied; `offline', set the server offline;
nil, ask user.  If the server is not covered by Gnus agent, set the
server denied."
...)


>  Or is offline mode only used for
> agentized servers?

Offline is the nnagent backend.  An unagentized server that fails to
connect is denied.

Kevin