Windows and UTF-8 in the filesystem

Sunday, January 4, 2004

I just tried playing some MP3s on my machine via a Windows 2000 machine across the room that has big speakers attached. WinAmp could play most but couldn’t access a couple. I checked the permissions and they were all right. Then I realised the file paths had non-ASCII characters in them (I was trying to play Noir Désir and Afro Celt Sound System’s Éireann), and Mac OS X’s HFS filesystem stores names in UTF-8, while Windows uses Latin-1. The tab key completion in bash says the name is “Noir De\314\201sir”, which looks like UTF-8 to me.

In the Windows file open dialog, “Noir Désir” appears correctly in the file listing, but as “Noir De´sir” in the path drop-down. I can add the file to the playlist in WinAmp, but WinAmp can’t access the file. I thought the problem may be with Samba, but smbclient can access it just fine.

Does this mean that Windows 2000 can’t handle Unicode in the filesystem?

bluesmoon — Jan 4, 2004 6:06:22 AM — # ↩

could also just be a problem with winamp storing strings as signed chars.
- Kiran Jonnalagadda — Jan 4, 2004 6:59:49 AM — # ↩
  
  Yes, it appears to be a WinAmp problem; Windows Media Player can open it.
  
  But the following are noteworthy:
  
  Exhibit A: Observe the window title
  
  Exhibit B: Observe the folder drop-down
evan — Jan 4, 2004 6:52:45 PM — # ↩

the windows apis are split between the non-unicode and unicode versions. everything that uses the old api probably won't do stuff right. it makes sense that winamp wouldn't have the latest unicode stuff.