[cvsnt] Filename charset

Tony Hoyle tmh at nodomain.org
Wed Feb 25 10:16:20 GMT 2004


Robin Rosenberg wrote:

> Not a problem in practice since I think "international" servers would not
> use national characters. So even if all characters do not work, so be it, if
> the common ones work. It's really only a few that I'm interested in, i.e. åäöÅÄÖ,
> i.e. characters people tend to use.

It's a problem because eg. there are quite a few Japanese users of CVSNT 
- it's not really a solution if it excludes them.


> Have there been any discussions on how, i.e. an agreement in principle
> on how to solve this, so that if someone implements it or part of it patches
> won't be rejected right away?
> 
> I had a kludgy idea of setting up server on different ports. 2401 for ISOLatin1, 2402
> for UTF-8, but I'd rather skip that. It's a future problem, but it blocks any linux client
> from having clients with UTF-8 as the default charset. Som linux distributions nowadays
> default to UTF-8.

... which is a problem because NT support for UTF8 is minimal (in fact, 
it's virtually nonexistant, beyond converting to/from Unicode).  Since 
all the text output of CVS is generated by the server, it impacts almost 
every area of the code if there's going to be any changes.

How do eg. Sourceforge handle this?

I have tried to produce a UTF8 version of CVSNT and decided the effort 
at the moment outweighs the advantages - there's the new code coming 
onstream that'll be UTF8 at the core (the first chunk of that is due to 
go in once the current lot becomes stable) but that's a long process and 
will only happen in stages (long = 6 months if the funding happens).

Tony



More information about the cvsnt mailing list