Hi,

On Wed, 04 Feb 2009 15:56:03 +1100, Tanaka Akira <akr / fsij.org> wrote:

> Fixed length characters needs variable length bytes in multibyte  
> encoding.

That's right.

> So the record is variable length in byte layer.  Is it ok?

Yes - that's why I want to be able to read N chars, because I don't know  
how many bytes there are!

> In such format, record-wise update cannot be implemented efficiently.

That's right - These files are quite small, and I only need to do  
sequential I/O. I want to keep the format backward compatible when using a  
single-byte encoding.

For the record, I didn't request character-based seek - my comment saying  
they couldn't be implemented efficiently was replying to someone else who  
suggested it.

Cheers
Mike