Bazaar internationalization

[John Arbash Meinel]

Alexander Belchenko wrote:
> Dennis Benzinger пишет:
>> I suggest adding a --encoding switch to Bazaar to specify in which locale
>> a commit message is encoded. If the switch isn't supplied on a commit then
>> the system encoding is used.
> 
> Well, bzr store commit messages in utf-8 internally without any special flag.
> And still no one complains that his log is broken.
> So I see only one reason for this flag: when commit messages come from text file
> (-f option). So yes, this option could be useful.
> 
>> This is helpful for example in international projects if the project
>> language is chinese and an english speaking developer has to write a commit
>> message in chinese. With this switch he can do that without changing his
>> system encoding.
> 
> I don't understand how someone could write chinese messages with ascii terminal.
> Please give me example.
> 
> Many Linux terminals today used utf-8 (AFAIK). So you don't need to switch anything.
> Things worse on win32, but I don't understand eastern languages, so I just
> cannot say anything valuable.
> 
>> (This idea was taken from Subversion. See:
>> http://svnbook.red-bean.com/nightly/en/svn.ref.svn.html)
> 
> I think this needed for successful cvs2svn conversion as "must have" thing.

I think all converters that I've used forgo trying to do things through
'bzr' and instead directly hook into 'bzrlib'. This lets them work in
terms of Unicode strings, and gives easy ways to directly set the
timestamp, etc.

So we don't have quite the same requirements that you can configure
everything from the command line, because you can do a lot of it
internally. This simplifies the command line for the general person (who
should not be changing the timestamp or explicit revision ids of their
commits), but exposes the functionality to code that does need it.

That said, having a way to specify the encoding of a file is probably
important. I think it can be done with just:

LANG=whatever bzr commit -F text.txt

But that is a bit of a hack.

So in general, I'm not opposed to adding '--encoding', though I would
generally recommend waiting until it is explicitly needed, rather than
assuming it is already needed.
> 
>> Should there be a wiki page where the internationalization blueprint can be
>> improved?
> 
> Yes, it should. I create a stub: http://bazaar-vcs.org/I18nSupport
> Feel free to start fill it at least minimally.
> 
> [µ]

John
=:->