Do we still do Incoming* files?

Do we still do Incoming* files?

[lconrad]

One of my problems recovering from the hæmorraging cvs-gnus last
monday, which in the heat of the moment I failed to report, was that in 
addition to putting my mail in the wrong place, it also hadn't
written it to its usual Incoming* file.  

The mail is now going to the right place, but it still doesn't seem to 
be writing Incoming* files.  Is this a bug or a feature?

I would say based on what I felt like Monday morning that it's a bug.

-- 
Laura (mailto:lconrad@world.std.com , http://www.world.std.com/~lconrad/ )
(617) 661-8097	fax: (801) 365-6574 
233 Broadway, Cambridge, MA 02139

Re: Do we still do Incoming* files?

[Kai Großjohann]

lconrad@world.std.com writes:

> The mail is now going to the right place, but it still doesn't seem to 
> be writing Incoming* files.  Is this a bug or a feature?

What's the value of mail-source-delete-incoming?  If that's nil and
there are no Incoming files, we're talking about a serious bug.  Else,
it's not too bad.  Maybe CVS users should be setting that variable
themselves, anyway?

kai
-- 
~/.signature: No such file or directory

Re: Do we still do Incoming* files?

[Karl Kleinpaste]

lconrad@world.std.com writes:
> The mail is now going to the right place, but it still doesn't seem to 
> be writing Incoming* files.  Is this a bug or a feature?

mail-source-delete-incoming is default nil in [a-z]gnus-0.X releases,
but default t once released as 5.Y, the rationale being that alpha-
testers are liable to hurt themselves and need to be able to recover,
while those using a released version ought not be bothered with such
stuff.

Re: Do we still do Incoming* files?

[lconrad]

>>>>> "Kai" == Kai Großjohann <Kai.Grossjohann@CS.Uni-Dortmund.DE> writes:

    Kai> lconrad@world.std.com writes:
    >> The mail is now going to the right place, but it still doesn't seem to 
    >> be writing Incoming* files.  Is this a bug or a feature?

    Kai> What's the value of mail-source-delete-incoming?  If that's nil and
    Kai> there are no Incoming files, we're talking about a serious bug.  

It's t now.  It used to be nil, and I didn't change anything.  

    Kai> Else, it's not too bad.  Maybe CVS users should be setting
    Kai> that variable themselves, anyway?

Yes, I agree.   But when a default like that changes, it would be good 
to let people know.


-- 
Laura (mailto:lconrad@world.std.com , http://www.world.std.com/~lconrad/ )
(617) 661-8097	fax: (801) 365-6574 
233 Broadway, Cambridge, MA 02139

Re: Do we still do Incoming* files?

[Karl Kleinpaste]

lconrad@world.std.com writes:
> Yes, I agree.   But when a default like that changes, it would be good 
> to let people know.

`C-h i' -> Appendices -> History -> Gnus Development

"   Some variable defaults differ between alpha Gnusae and released
"Gnusae.  In particular, `nnmail-delete-incoming' defaults to `nil' in
"alpha Gnusae and `t' in released Gnusae.  This is to prevent lossage of
"mail if an alpha release hiccups while handling the mail.

Re: Do we still do Incoming* files?

[Kai Großjohann]

Karl Kleinpaste <karl@charcoal.com> writes:

> [...] the rationale being that alpha- testers are liable to hurt
> themselves and need to be able to recover,[...]

It seems that gnus fresh from CVS might also benefit from some
protection, though.  Viz, the missing-slash problem of a couple of
days ago.

kai
-- 
~/.signature: No such file or directory

Re: Do we still do Incoming* files?

[Karl Kleinpaste]

Lars will have to be convinced to turn on -delete-incoming only when
creating released tarballs, so that it is always nil in CVS.

I'm not convinced that this is a Good Thing.

I tend to think that anyone choosing deliberately to work with any
version under development, rather than one of those released tarballs,
would have to choose as well whether to defend himself from this class
of problem.

Re: Do we still do Incoming* files?

[Kai Großjohann]

Karl Kleinpaste <karl@charcoal.com> writes:

> I tend to think that anyone choosing deliberately to work with any
> version under development, rather than one of those released tarballs,
> would have to choose as well whether to defend himself from this class
> of problem.

Maybe you're right.  Hm.

Yes, users of CVS should know to set mail-source-delete-incoming to
nil themselves.

Users of [srpo]gnus-*.tar.gz will have that done for them.

Users of gnus-*.tar.gz will have mail-source-delete-incoming to t.

I think we can agree on that.  (Note how I carefully mentioned that
Gnus might benefit from the protection, but not who does the
protecting :-))

kai
-- 
~/.signature: No such file or directory

Re: Do we still do Incoming* files?

[lconrad]

>>>>> "Kai" == Kai Großjohann <Kai.Grossjohann@CS.Uni-Dortmund.DE> writes:

    Kai> Yes, users of CVS should know to set mail-source-delete-incoming to
    Kai> nil themselves.

    Kai> Users of [srpo]gnus-*.tar.gz will have that done for them.

    Kai> Users of gnus-*.tar.gz will have mail-source-delete-incoming to t.

    Kai> I think we can agree on that.  

OK, but in that case the instructions for using cvs should mention
it.  I feel less protected using cvs than I did using pgnus tarballs.
If I waited a day to download a pgnus tarball, I was pretty sure that
someone had used it before and would have mentioned any terrible
problem.  With cvs, I have no idea whether I'm the first user or
not.

So given I have thought about it, I have no problem with setting the
variable for myself. But I think it's a good idea to make sure that
people think about it.


-- 
Laura (mailto:lconrad@world.std.com , http://www.world.std.com/~lconrad/ )
(617) 661-8097	fax: (801) 365-6574 
233 Broadway, Cambridge, MA 02139

Re: Do we still do Incoming* files?

[Kai Großjohann]

lconrad@world.std.com writes:

> OK, but in that case the instructions for using cvs should mention
> it.  I feel less protected using cvs than I did using pgnus
> tarballs.  If I waited a day to download a pgnus tarball, I was
> pretty sure that someone had used it before and would have mentioned
> any terrible problem.  With cvs, I have no idea whether I'm the
> first user or not.

Well, with CVS you are, at worst, the second user.  The next step
would be to use OVC (Oracle Version Control) which allows you to use
the modifications before the developer makes them.  Then, you will
_really_ feel to be living on the bleeding edge!  ;-)

kai
-- 
~/.signature: No such file or directory

Re: Do we still do Incoming* files?

[lconrad]

>>>>> "Kai" == Kai Großjohann <Kai.Grossjohann@CS.Uni-Dortmund.DE> writes:

    Kai> Well, with CVS you are, at worst, the second user.  

I am? 

Is this because the developers test before committing, or because whoever
commits to cvs tests before releasing to the rest of the world?

And what do they test?  

In any case, I doubt that I was ever even as early as the 10th user of 
pgnus.  I do remember one that had some pretty major problem, and
several people in Europe had reported it between the time Lars
announced it and the time I woke up in the morning.  So I didn't
download that one.

-- 
Laura (mailto:lconrad@world.std.com , http://www.world.std.com/~lconrad/ )
(617) 661-8097	fax: (801) 365-6574 
233 Broadway, Cambridge, MA 02139

Re: Do we still do Incoming* files?

[Gunnar Evermann]

lconrad@world.std.com writes:

> In any case, I doubt that I was ever even as early as the 10th user of 
> pgnus.  I do remember one that had some pretty major problem, and
> several people in Europe had reported it between the time Lars
> announced it and the time I woke up in the morning.  So I didn't
> download that one.

why do you use CVS then? stick with the tarballs or use the CVS tags
to checkout snapshots/releases, i.e. 5.8.3 atm.

 Gunnar

Re: Do we still do Incoming* files?

[Kai Großjohann]

lconrad@world.std.com writes:

> >>>>> "Kai" == Kai Großjohann <Kai.Grossjohann@CS.Uni-Dortmund.DE> writes:
> 
>     Kai> Well, with CVS you are, at worst, the second user.  
> 
> I am? 

You are, unless you're not because the developer didn't pay
attention.  Such are the joys of CVS ;-)

SCNR mentioning the OVS thing in that message :-)  I think I've been
reading too much Terry Pratchett.

kai
-- 
~/.signature: No such file or directory

Re: Do we still do Incoming* files?

[lconrad]

>>>>> "Kai" == Kai Großjohann <Kai.Grossjohann@CS.Uni-Dortmund.DE> writes:

    Kai> Well, with CVS you are, at worst, the second user.  
    >> 
    >> I am? 

    Kai> You are, unless you're not because the developer didn't pay
    Kai> attention.  Such are the joys of CVS ;-)

Where I hang out, one of the major things testers test is whether
developers are paying attention.  We find they sometimes don't.

-- 
Laura (mailto:lconrad@world.std.com , http://www.world.std.com/~lconrad/ )
(617) 661-8097	fax: (801) 365-6574 
233 Broadway, Cambridge, MA 02139

Re: Do we still do Incoming* files?

[Harry Putnam]

A slightly different angle on this subject: 

I use even more thorough backup than gnus affords.  Worrying
about what gnus might do to itself or your mail is one thing, but I've
found that I tinker with gnus more than any other piece of software I
use.

Trying different things out, experimenting with the advanced features
etc., so in my mind there is another substantial risk of what I might
inadvertantly do through ill informed tinkering.  Probably far
outweighing any developer foibles that might come up.

I realize that most of the users on this list are probably too skilled
to do harmfull things to there mail, but I have inadvertantly leveled
both ~/Mail and ~/News on more than one occasion.  Brought about buy
tinkering with gnus.  In that case, Incoming files won't help... he
he.

I'd recommend any linux/unix users should have a procmail recipe to
backup all incoming mail to a directory unrelated to gnus.

I have a  procmail recipe and a shell script that keep the most recent
600 incoming messages.  The script is just a little "sed" script that
deletes anything over 600 files, like:

   rm `ls -t /home/reader/spool/backup/[0-9]* | sed -e 1,600d` 2> /dev/null

And have procmail duplicate *all* incoming mail to ~/spool/backup/
in one message per file format.

   :0 c
   backup/.

I run the sed script whenever the spirit moves me... So I really
usually have something like 1,000 of the most recent messages.

I keep the variable to delete Incoming set to "t" as a matter of
course because of the safer technique detailed above.

I also keep a "nndir" group of the backup directory in case I want to
respool something or just need to have a look for one reason or
another.

I've found it to be quite handy having that backup directory on many
many occasions.  It does cost a little disk space, but with a single
user machine that has 30 Gigs.... who cares?

Re: Do we still do Incoming* files?

[Lars Magne Ingebrigtsen]

Karl Kleinpaste <karl@charcoal.com> writes:

> Lars will have to be convinced to turn on -delete-incoming only when
> creating released tarballs, so that it is always nil in CVS.
> 
> I'm not convinced that this is a Good Thing.

Mee too.  I've now changed the default.

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

Re: Do we still do Incoming* files?

[Laura Conrad]

>>>>> "Gunnar" == Gunnar Evermann <ge204@eng.cam.ac.uk> writes:

    Gunnar> why do you use CVS then? stick with the tarballs or use
    Gunnar> the CVS tags to checkout snapshots/releases, i.e. 5.8.3
    Gunnar> atm.

I'm not complaining about having hit the problem -- it made me feel
useful to be the first to report something like that.  

But I had thought I knew where the mail would be if gnus wasn't
finding it.  

I'm proposing a change in the documentation, so that other people may
be less confused than I was.

Right now, the instructions for using cvs say:

    gnus.org> The proper incantation for checking out the Gnus sources
    gnus.org> the first time is this:  

    gnus.org> ...    

    gnus.org> (You type the bold parts.) 

    gnus.org> You can then rename the resulting gnus directory to
    gnus.org> whatever you want -- like, for instance, pgnus.  

    gnus.org> After doing this, to get a fresh Gnus, all you have to
    gnus.org> do is say:  

    gnus.org> ...
    

    gnus.org> Please note that what you're getting is probably not even a
    gnus.org> bleeding-edge Gnus, but a severly hemorrhaging one. 

I'm proposing that we add something like:

    proposal> It's also a copy of what will be the released version,
    proposal> when it gets that far.  So the version number is gnus
    proposal> x.x.x, not <x>gnus-0.xx, and in order to report a
    proposal> problem, you need to specify the time you downloaded the
    proposal> cvs snapshot.  

    proposal> Also, an important protection for users of <x>gnusae is
    proposal> not in effect in the cvs repository.  That is, the
    proposal> variable mail-source-delete-incoming is set to t in the
    proposal> released versions, and in the CVS repository.  When an
    proposal> <x>gnus tarball is rolled, the default for this variable
    proposal> is changed to 'nil', so that a development version with
    proposal> a serious bug is less likely to lose people's mail.  So
    proposal> you probably want to think about the appropriate setting
    proposal> for this variable yourself.

This might reduce the frequency of the Frequently Asked Question "What 
are all these 'Incoming*' files in my ~/Mail directory", too.

If there are other differences between cvs gnus and the pgnus, or
future ognus, tarballs, that I don't know about yet, maybe someone
would like to write those up.  

-- 
Laura (mailto:lconrad@world.std.com , http://www.world.std.com/~lconrad/ )
(617) 661-8097	fax: (801) 365-6574 
233 Broadway, Cambridge, MA 02139