l10n approach for bzr

[Stephen J. Turnbull]

Barry Warsaw writes:
 > -----BEGIN PGP SIGNED MESSAGE-----
 > Hash: SHA1
 > 
 > On Mar 17, 2008, at 4:19 PM, John Arbash Meinel wrote:
 > 
 > > Of course, then you need to work out a numbering scheme that
 > > scales in a distributed system so that you don't get conflicts
 > > with 2 people adding different strings and calling them both
 > > 564. :)
 > 
 > Good point. :)

There is a standardized message ID system (IIRC standard is POSIX,
original vendor was Sun), but it's been so long since I did the
academic exercise of looking it up that I've forgotten its name.

And it *is* an academic exercise to look to it up; translators hate it
if you provide the number with the original string, because of course
they're expected to verify it.  If you hide it, what good does it do?
Programmers are similarly uncool with the idea, because they have to
keep an updated list of "used message ids", and worse, check it.

The best solution I know of, keeping programmers and translators
reasonable happy, is to use the fuzzy translation facilities, which
consider two strings to be probably the same if they're close in some
"minimum edit metric".  They work well.

As pointed out Launchpad already implements this well.