[RFC] nnmbox and first-time users

[RFC] nnmbox and first-time users

[Didier Verna]

        The nnmbox backend uses ~/.mbox by default. IMHO, it should have used
something more gnusy (or is there a rationale behind this value ?), at least
honoring gnus-directory, and probably message-directory. Obviously, I
understand that such an important variable cannot have its default changed
like that.

        Still, there might be interference problems with other software that
use ~/.mbox (like mutt). I've done a quick test with an mbox already filled
with one message by mutt, and starting a first-time gnus session. The message
is not erased, but will not be seen by gnus. That can be disconcerting and I
suspect much more serious stuff can happen with other software / other
situations.

        What if we made first-time gnus check if nnmbox is used, if the mbox
file already exists, and in that case, issue a warning, advice to move it away
and abort ?

-- 
Didier Verna, didier@lrde.epita.fr, http://www.lrde.epita.fr/~didier

EPITA / LRDE, 14-16 rue Voltaire   Tel.+33 (1) 53 14 59 47
94276 Le Kremlin-Bicêtre, France   Fax.+33 (1) 53 14 59 22   didier@xemacs.org

Re: [RFC] nnmbox and first-time users

[Kai Großjohann]

Didier Verna <didier@lrde.epita.fr> writes:

> Obviously, I understand that such an important variable cannot have
> its default changed like that.

In fact I think that slowly the time might have come to change a lot
of important variables in an incompatible manner.  That will at least
show the users that something is wrong, and so they can correct it.

We need to change the files and directories used for the servers to
something unique.  For example, use ~/Gnus/nnml/unnamed/ for the
server (nnml ""), and so on.

But there is a problem with the old .newsrc.eld files.  Using the old
file with the new directory is going to be a problem, but if we use a
new file then it's difficult for users to resurrect marks.

Thoughts?

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

Re: [RFC] nnmbox and first-time users

[Paul Jarc]

Kai.Grossjohann@CS.Uni-Dortmund.DE (Kai Großjohann) wrote:
> We need to change the files and directories used for the servers to
> something unique.  For example, use ~/Gnus/nnml/unnamed/ for the
> server (nnml ""), and so on.

That would (I assume) interfere with a server called (nnml "unnamed").
Instead, we could use ~/Gnus/nnml/ for (nnml "") and
~/Gnus/nnml+unnamed/ for (nnml "unnamed").  Or something like that.

(nnml "foo") would still interfere with (nnml "foo/bar").  Should we
worry about such cases?  Slashes in server addresses could be encoded
in directory names somehow, but I don't know if it's worth the effort.
How likely is it that someone would use such an address, and also not
specify the nnml-directory server parameter?

> But there is a problem with the old .newsrc.eld files.  Using the old
> file with the new directory is going to be a problem,

How so?


paul

Re: [RFC] nnmbox and first-time users

[Christoph Conrad]

    > In fact I think that slowly the time might have come to change a
    > lot of important variables in an incompatible manner. That will
    > at least show the users that something is wrong, and so they can
    > correct it.

If that is done this change should be documented in a VERY explicit
manner accompanied by a good "transition description".

Re: [RFC] nnmbox and first-time users

[Kai Großjohann]

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

> Kai.Grossjohann@CS.Uni-Dortmund.DE (Kai Großjohann) wrote:
>> We need to change the files and directories used for the servers to
>> something unique.  For example, use ~/Gnus/nnml/unnamed/ for the
>> server (nnml ""), and so on.
>
> That would (I assume) interfere with a server called (nnml "unnamed").
> Instead, we could use ~/Gnus/nnml/ for (nnml "") and
> ~/Gnus/nnml+unnamed/ for (nnml "unnamed").  Or something like that.

Good idea.

> (nnml "foo") would still interfere with (nnml "foo/bar").  Should we
> worry about such cases?  Slashes in server addresses could be encoded
> in directory names somehow, but I don't know if it's worth the effort.
> How likely is it that someone would use such an address, and also not
> specify the nnml-directory server parameter?

Whee.  Until recently, we wouldn't have had to worry, since "/" was
forbidden in group names.  But now there is IMAP and now "/" in group
names is possible.

Ick.

>> But there is a problem with the old .newsrc.eld files.  Using the old
>> file with the new directory is going to be a problem,
>
> How so?

Because the new directory is empty (initially).

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

Re: [RFC] nnmbox and first-time users

[Kai Großjohann]

Christoph Conrad <christoph.conrad@gmx.de> writes:

>     > In fact I think that slowly the time might have come to change a
>     > lot of important variables in an incompatible manner. That will
>     > at least show the users that something is wrong, and so they can
>     > correct it.
>
> If that is done this change should be documented in a VERY explicit
> manner accompanied by a good "transition description".

Yes.  The Gnus startup code should have some facility to detect the
situation, alert the user to it, and DTRT.

Hm.  Coming to think of it, maybe it is better not to change the
default directories, but rather to have Gnus perform some sanity
checks.

So when Gnus discovers that the user has two servers pointing to the
same directory, Gnus would then barf.

Ah.  We discussed about this a while ago.  Sorry, I completely
forgot.  Yes, I think the sanity check approach is much better.

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