Re: arch, svn, cvs

[Tollef Fog Heen <tfheen@err.no>]

* Peter Samuelson 

| [Tollef Fog Heen]
| > I just stumbled across one issue: it doesn't handle the case where
| > you change your encoding without checking out the repository again:
| > 
| > : tfheen@thosu ~/svn/trunk > LANG=en_US.UTF-8 svn st
| > svn: Valid UTF-8 data
| > (hex: 46)
| > followed by invalid UTF-8 sequence
| > (hex: f8 72 73 74)
| 
| That's a design decision - to recode filenames to a user's locale.
| This enables users with different locales to all see the same filenames
| in their local checkouts.  You can argue that it's good or bad, but the
| alternative is *not* to recode them, and that causes exactly the same
| problem, when users of the same repository use different locales.  So
| when local filenames are in the "wrong" locale (for whatever reason),
| what *should* the software do?

If it does decide to check out using the local encoding, it could at
least have taken notice of this so «svn st» wouldn't later just fail.

| Or, another point: most software does not keep an internal database of
| the encoding used for its data filenames.  Do you think subversion
| should?

If it tries to handle encodings in file names?  Yes.  If it goes a
more simplistic route and just treats file names as arrays of bytes?
No.  Either of those solutions is fine, but somewhere in between is
not.

-- 
Tollef Fog Heen                                                        ,''`.
UNIX is user friendly, it's just picky about who its friends are      : :' :
                                                                      `. `' 
                                                                        `-