Re: [O] [RFC] Change some defcustoms into defcont

[Carsten Dominik]

Hi Nicolas, and everyone else.

I have just investigated what actually happens is someone has a
customize setting for a variable in his/her setup, and then the variable
is changed to a defconst  behind his/her back.  The result was very
surprising to me:  customize-set-variables still does set the variable
to the stored value.  Changing a different customize variable
and selecting "apply and change" updates the init file where
customize-set-variables is stored, and interestingly still preserves
the customize setting for the variable that now has become
a defconst.

This surprised me, and it seems to indicate that existing configurations
would not break.

Furthermore, I have looked at the old customization survey
at http://orgmode.org/worg/org-configs/org-customization-survey.html

It turns out that none of the variables Nicolas listed seems to have been
customized by the people participating in this survey.

Tanking these together, I am now inclined to change my view and
to let Nicolas make his proposed change.
What I would like to request is that we try to make Org still work if these
variables do in fact have a different value - but that is what Nicolas
already said.

So:  IF YOU HAVE CUSTOMIZED any of this variables, I'd like to hear about it.
Otherwise, I now think that the proposal is actually good and should move ahead.

- Carsten


On Oct 23, 2013, at 10:56 AM, Nicolas Goaziou <address@hidden> wrote:

> Hello,
> 
> Carsten Dominik writes:
> 
>> can you remind me what the bug was?  The taskjuggler issue?
> 
> Correct. I will fix it today.
> 
>> Yes, as I said, I do see all these problems, but I also feel the 
>> responsibility
>> to break as few as possible existing configurations.
> 
> I understand. Though, almost any serious change breaks some existing
> configuration. It is unfortunate, but I hope it will not prevent us from
> correcting past, and future, mistakes. Without that right to be wrong,
> we'd better plan Org development a very long time ahead and cross
> fingers hoping we will never fail. Don't cross too hard.
> 
>> If you want, I can take a shot at documenting this properly.
> 
> I'm not convinced by that documentation thing, i.e. "we allow you to
> change the string, but we remind you that you shouldn't do it". It even
> goes against my initial suggestion, so I cannot say I "want" anything in
> that area.
> 
> Unless, that is, you're talking about a deprecation warning, e.g. "In
> future releases, this variable will not be customizable anymore, so be
> careful when changing its default value". But, IIUC, that's not what you
> want.
> 
> 
> Regards,
> 
> -- 
> Nicolas Goaziou