> On Fri, Jul 21, 2006 at 07:10:20PM +0100, David Laight wrote:
> > So you continue to export fhandle_y, but don't expose the header file
> > that defines any contents.

i'm not sure if it's a good idea given the typical usage of these api.
ie. handle filehandle from/to network.
anyway, we can change it later because it makes no abi difference.

> Hmmm, we could actually define a special version (completely independend
> of the kernel fhandle_t) for userland:
> 
>   typedef struct fhandle {
>       size_t fh_size;
>   } fhandle_t;

i don't think it is a good idea, because it isn't compatible with
what kernel uses.  you'll end up to convert between them in mountd etc.

YAMAMOTO Takashi