Re: Emacs i18n

[Eli Zaretskii]

> From: Juri Linkov <address@hidden>
> Cc: address@hidden,  address@hidden,  address@hidden
> Date: Fri, 08 Mar 2019 00:29:17 +0200
> 
> > We still should try to use .po files for them, if at all possible,
> > and perhaps also the gettext code that supports looking up strings
> > in .gmo catalogs generated from .po.
> 
> The PO format is best suited for translation of one-liners like
> messages and menu items, but I doubt that the PO format would be
> the most efficient implementation for multi-line doc strings since
> gettext uses the whole text of the doc string as a key to translation.

I'm not sure I understand why the length of the string is an important
factor here.  Can you explain?  If the problem is with the efficiency
of gettext implementation of indexing, then we could have our own
indexing method.

> Whereas more efficient would be to use a Lisp symbol (function or
> variable name) as a translation key.

A key other than the original string would mean abandoning the PO
format.  Any deviation from PO would mean major PITA for translation
teams, so we should make sure the reason for such a deviation is a
very good reason.  I'm not yet sure we have such a good reason.