> Hi Bill,
>
> On Wed, Dec 31, 1997 at 10:08:46AM -0500, Bill Hawes wrote:
> > Unless someone else is contemplating changes, I think it's safe to
> > declare the kernel API for knfsd is stable. There undoubtedly are bugs
> > remaining to be fixed, but this shouldn't require any changes to the
> > API, and cleaning up the include files and linux-nfs package will
> > encourage more people to test knfsd.
>
> You're right. What's more, if there are changes to the API I included
> an API version number in all argument structs which we can use if it
> changes. I'll try to build a new version RSN.
>
> The only thing that's worrying me are possible discrepancies between kernel
> and user land types such as uid_t and dev_t. If they don't match, we'll
> have a serious problem. Maybe we need to use generic types in the argument
> structs (e.g. unsigned int instead of uid_t). Any comments on this?
That's my vote. If you look at the patches I sent you for the Alpha
(crude though they may be), this is the effect of what I did on the user
side.
Steve