Re: Q: something like autoload for coding-systems?

[Eli Zaretskii]

On Sun, 4 Nov 2001, Stefan Monnier wrote:

> > I'm not sure it's actually worth autoloading a few tens of 8-bit
> > coding systems -- I haven't measured how much space they use.
> 
> But if we want to be able to encode/decode things like big chinese
> charsets into Unicode, it might being to make sense to autoload those
> big tables (assuming the mapping between the two is such a big table
> and can't be easily compressed).

I agree with both of you ;-)

The reason that cpNNN coding systems are not preloaded is mainly
because they were originally supposed to be of importance on
DOS/Windows systems only.  At the time I wrote codepage.el, it seemed
wrong to bloat the Unix executable with something which will rarely if
ever be used, even if the size was not so large.

If that assumption is no longer true, perhaps we should always preload
those coding systems.

I also agree with Stefan that Unicode support will require large
tables to be loaded on demand, so maybe we should have some machinery
to do that.

Note that Unicode-related tables are needed not only for code
conversions, but also for defining character properties, such as case
conversion, bidi categories, syntax classification, etc.  We currently
have these for a small number of special cases, mostly for non-Unicode
characters, but a more general solution is probably required.

(The last issue is related to something we once discussed: what data
structures do we need to represent such information efficiently within
Emacs.)

> Also we could even make *all* coding systems autoloaded, at which point
> all those 8bit tables can start to accumulate into something more
> measurable.

Sounds like something we should consider.