Re: ChangeLog and unification

[Eli Zaretskii]

On Mon, 18 Feb 2002, Stefan Monnier wrote:

> The unify-on-encode is safe and does not do unification.

What will it do if you yank a string with Latin-2 characters and then 
save the buffer?  Assuming that buffer-file-coding-system is 
iso-2022-7bit, will it save those Latin-2 characters as Latin-2 or as
Latin-1?  Also, what if I type characters with a latin-2 input method,
or with keyboard-coding-system set to latin-2--will the characters be 
saved as Latin-2 or Latin-1?

> All it does is to allow saving latin-1 chars into a latin-2 file (as long
> as the latin-1 chars also exist in latin-2, of course).

In a file encoded with one of the iso8859-x encodings, this is indeed a 
no-op.  The plot might thicken when you use iso-2022-derived encodings, 
which tag characters with their charset id.  (I say ``might'' because I 
really don't know what happens: I didn't yet have time to give 
unify-on-* modes a serious test ride.)

> I really think unify-on-encode should be turned ON by default
> because I still haven't heard of any reasonable scenario where it can
> do anything undesirable.

You may be right, but the bitter experience with Mule-related issues 
teaches me that we should be extra-cautious with turning features on by 
default.

> Actually, I can't even think of any reason
> why we should try to make it possible to turn it off.

That is almost certainly a bad idea: it should be possible to turn off
_any_ feature, certainly a Mule-related one.  Someone, some day, will 
want that.