Re: Emacs 26.1 release branch created

[Alan Mackenzie <acm@muc.de>]

Hello, Eli and John.

I'm putting your two posts together to answer them together.

On Fri, Sep 22, 2017 at 10:17:15 +0300, Eli Zaretskii wrote:
> > Date: Thu, 21 Sep 2017 20:54:47 +0000
> > From: Alan Mackenzie <acm@muc.de>
> > Cc: Richard Stallman <rms@gnu.org>

> > I have pushed a first draught into the new branch
> > scratch/customize-quotes.  Comments would be welcome.

> > One thing which I hope isn't too controversial, is that I have redefined
> > the value of nil in text-quoting-style to mean "no translation of
> > quotes" and introduced t to mean "prefer curved quotes", the meaning nil
> > previously had.  The default value for text-quoting-style remains
> > "prefer curved quotes".

> This _is_ controversial, since it means we are introducing a backward
> incompatible change.

> > My reasoning here was that having nil meaning "do nothing" is
> > intrinsically the Right Thing.  Also very few people, if any, will
> > already have explicitly set the variable to nil.  And in any case, the
> > use of the variable was restricted to experts, all of whom will be able
> > to adapt to the new nil and t without difficulty.

> IMO, this reasoning is too weak to justify the incompatible change.

On Thu, Sep 21, 2017 at 22:58:40 -0700, John Wiegley wrote:
> >>>>> "PE" == Paul Eggert <eggert@cs.ucla.edu> writes:

> PE> Let's not change the semantics of text-quoting-style's values.   Such a
> PE> change is not worth the compatibility hassle to users. John asked you to
> PE> propose a patch to make text-quoting-style customizable, not to change its
> PE> semantics. Whether text-quoting-style is customizable is orthogonal to
> PE> what its values mean, and we shouldn't conflate the two issues.

> Just to confirm, Paul's assessment is correct: at this late stage, I only want
> to add customizability, no other changes.

I'm surprised and dismayed by both your reactions to my proposed patch.
I thought the issues through carefully before investing time and energy
in it.

You talk about backward incompatibility.  Well, backward incomptibility
is only bad when it inconveniences users, which is most of the time.  It
isn't the case here.  Neither of you, nor Paul has suggested a scenario
in which a user might suffer.  I put it to you that there aren't any,
particularly considering how text-quoting-style was kept away from
ordinary users in Emacs 25.

The current set of symbols used for text-quoting-style is bad.  Nobody
in recent posts, if ever, has defended them as being good.  Its worth
bearing in mind that they found their way into the Emacs master without
any sort of review beforehand.  How are we going to defend the use of
the symbol `grave' having the meaning "don't do any translation" five or
ten years from now?  What makes it worse is its suggestion of "grave
danger" alongside "grave accent".

I'm concerned for the quality of Emacs, and the use of nil in
text-quoting-style sticks out like a sore thumb.  The only possible time
to fix this problem is now - after text-quoting-style is released for
ordinary users in Emacs-26, it will be too late.

So, I'm asking you once more to consider very carefully the issues
behind my proposed amendment.  For my part, I undertake to accept your
decision on the matter without further argument, and to change the code
in scratch/customize-quotes to the simple change you were expecting,
should you say no again.

Thanks.

-- 
Alan Mackenzie (Nuremberg, Germany).