Make Gnus emit more errors for unknown servers?

Make Gnus emit more errors for unknown servers?

[Kai Großjohann]

I think if Gnus finds a group name where no server exists, it just
creates a server on the fly.  So for example, if people add

    (nnml "" (nnml-directory "/tmp/foo"))

to gnus-secondary-select-methods, then create some groups in that
server, then accidentally delete that entry from
gnus-secondary-select-methods, Gnus will create a (nnml "") server on
the fly, and then it won't be able to open the groups.

I think it is better to emit an error message in such a case.

(Hm.  Though it is not clear how people are supposed to add the
missing server to the server buffer in such a case -- they can't
start Gnus.  Hm.  Ideas?)

kai
-- 
I like BOTH kinds of music.

Re: Make Gnus emit more errors for unknown servers?

[Lars Magne Ingebrigtsen]

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

> I think if Gnus finds a group name where no server exists, it just
> creates a server on the fly.  So for example, if people add
>
>     (nnml "" (nnml-directory "/tmp/foo"))
>
> to gnus-secondary-select-methods, then create some groups in that
> server, then accidentally delete that entry from
> gnus-secondary-select-methods, Gnus will create a (nnml "") server on
> the fly, and then it won't be able to open the groups.

The server should still be in gnus-server-alist, I think?  If it
isn't, it should be put there.  Probably.

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

Re: Make Gnus emit more errors for unknown servers?

[Kai Großjohann]

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

> Kai.Grossjohann@CS.Uni-Dortmund.DE (Kai Großjohann) writes:
>
>> I think if Gnus finds a group name where no server exists, it just
>> creates a server on the fly.  So for example, if people add
>>
>>     (nnml "" (nnml-directory "/tmp/foo"))
>>
>> to gnus-secondary-select-methods, then create some groups in that
>> server, then accidentally delete that entry from
>> gnus-secondary-select-methods, Gnus will create a (nnml "") server on
>> the fly, and then it won't be able to open the groups.
>
> The server should still be in gnus-server-alist, I think?  If it
> isn't, it should be put there.  Probably.

I'm trying to suggest the opposite: when Gnus finds that a server is
missing, it should issue an error message.

There have been discussions in the past of people who see lots of
strange servers when they do `^'.  And even more strange problems
resulted from this.

kai
-- 
Simplification good!  Oversimplification bad!  (Larry Wall)

Re: Make Gnus emit more errors for unknown servers?

[Lars Magne Ingebrigtsen]

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

>> The server should still be in gnus-server-alist, I think?  If it
>> isn't, it should be put there.  Probably.
>
> I'm trying to suggest the opposite: when Gnus finds that a server is
> missing, it should issue an error message.

But wouldn't it be better if the server never went missing?

> There have been discussions in the past of people who see lots of
> strange servers when they do `^'.  And even more strange problems
> resulted from this.

Well, it is valid in Gnus to just put arbitrary select methods as the
servers of arbitrary groups.  This will result in strange servers in
the server buffer.  Is that a problem?

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

Re: Make Gnus emit more errors for unknown servers?

[Kai Großjohann]

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

> Kai.Grossjohann@CS.Uni-Dortmund.DE (Kai Großjohann) writes:
>
>>> The server should still be in gnus-server-alist, I think?  If it
>>> isn't, it should be put there.  Probably.
>>
>> I'm trying to suggest the opposite: when Gnus finds that a server is
>> missing, it should issue an error message.
>
> But wouldn't it be better if the server never went missing?

No.  If people make a mistake in changing their server configs, then
Gnus adding the old server back again doesn't really help so much.

For example, in the old days, the select method entry in .newsrc.eld
was (nnml "") for each nnml group.  Now changing
gnus-secondary-select-methods to add some parameter for the server
spec won't do useful things.

Now Gnus puts the string "nnml:" there rather than (nnml ""), so that
it works to change the server parameters in
gnus-secondary-select-methods.  This is one step in preventing
strange servers from appearing.

Do you see the point I'm trying to make?  The user tries to change
something about a server, but something happens so that Gnus does not
use the new server definition for some groups.  Then my opinion is
that Gnus should warn the user that something is wrong.  But the
current behavior for Gnus is that Gnus silently adds the `missing'
server.

>> There have been discussions in the past of people who see lots of
>> strange servers when they do `^'.  And even more strange problems
>> resulted from this.
>
> Well, it is valid in Gnus to just put arbitrary select methods as the
> servers of arbitrary groups.  This will result in strange servers in
> the server buffer.  Is that a problem?

I want Gnus to allow the user to create strange servers on purpose,
but I also want Gnus to help the user from accidentally creating
strange servers.

kai
-- 
Simplification good!  Oversimplification bad!  (Larry Wall)

Re: Make Gnus emit more errors for unknown servers?

[Lars Magne Ingebrigtsen]

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

> Do you see the point I'm trying to make?  The user tries to change
> something about a server, but something happens so that Gnus does not
> use the new server definition for some groups.  Then my opinion is
> that Gnus should warn the user that something is wrong.  But the
> current behavior for Gnus is that Gnus silently adds the `missing'
> server.

Yes -- but my point is that a server shouldn't go missing.  :-)

Let's say you have the (nnml "" (...)) server in your secondary select
methods.  Your groups will put the "nnml:" server in the select method
slot, and everything works fine.  Then the user removes that server
from the secondary select methods, and Gnus then knows nothing about
how to select that group.

Wouldn't it be nice if that couldn't happen?  If Gnus were to store
the server in `gnus-server-alist' (or some other variable), then the
user could remove anything from the list of secondary servers, and
nothing would break.

Of course, the list of servers would have to be synced with the
secondary server list, etc.

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

Re: Make Gnus emit more errors for unknown servers?

[Paul Jarc]

Lars Magne Ingebrigtsen <larsi@gnus.org> wrote:
> Let's say you have the (nnml "" (...)) server in your secondary select
> methods.  Your groups will put the "nnml:" server in the select method
> slot, and everything works fine.  Then the user removes that server
> from the secondary select methods, and Gnus then knows nothing about
> how to select that group.
>
> Wouldn't it be nice if that couldn't happen?  If Gnus were to store
> the server in `gnus-server-alist' (or some other variable), then the
> user could remove anything from the list of secondary servers, and
> nothing would break.

You're talking about defending against modifying
gnus-secondary-select-methods within a Gnus session, right?  I think
Kai is talking about editing the setting between sessions.  The next
time you start Gnus, and Gnus finds a group whose server is missing,
it should not be automatically created, because Gnus doesn't have
enough information to do it correctly.

Or do you mean that secondary servers should be stored in .newsrc.eld
just in case they go missing?  In that case, when would a server ever
be removed from there?


paul

Re: Make Gnus emit more errors for unknown servers?

[Lars Magne Ingebrigtsen]

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

> Or do you mean that secondary servers should be stored in .newsrc.eld
> just in case they go missing?

Yes.  Or rather, all servers that Gnus has heard of should be stored
there. 

> In that case, when would a server ever be removed from there?

It wouldn't. 

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

Re: Make Gnus emit more errors for unknown servers?

[Paul Jarc]

Lars Magne Ingebrigtsen <larsi@gnus.org> wrote:
> prj@po.cwru.edu (Paul Jarc) writes:
>> Or do you mean that secondary servers should be stored in .newsrc.eld
>> just in case they go missing?
>
> Yes.  Or rather, all servers that Gnus has heard of should be stored
> there.

And the server parameters in gnus-secondary-select-methods and such
would override those in .newsrc.eld, right?  Ok.

>> In that case, when would a server ever be removed from there?
>
> It wouldn't.

Then we'd be leaking a bit of disk space.  What if during startup, we
remove a select method from gnus-server-alist if it does not appear in
any variable like g-s-s-m, and no group info refers to it?  Or if that
would risk losing a foreign server, then we can just put these saved
select methods in a different variable in .newsrc.eld - one which
contains no foreign servers.


paul

Re: Make Gnus emit more errors for unknown servers?

[Lars Magne Ingebrigtsen]

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

> And the server parameters in gnus-secondary-select-methods and such
> would override those in .newsrc.eld, right?  Ok.

Yup.

> Then we'd be leaking a bit of disk space.  What if during startup, we
> remove a select method from gnus-server-alist if it does not appear in
> any variable like g-s-s-m, and no group info refers to it?  Or if that
> would risk losing a foreign server, then we can just put these saved
> select methods in a different variable in .newsrc.eld - one which
> contains no foreign servers.

Sure.  We can go through the newsrc-alist and adjust the saved list of
servers according to what we find there.

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