It looks as if I had resolved most of the issues - at least I can read
  from and write to coda-volumes on my sparc64-System.

That's very good news.  I look forward to seeing your patches (and
having them applied).

  - is there a reason why the CodaFid-member of some structs is usually
  called "Fid" in kernel-coda.h and "VFid" in userspace-coda.h? This is
  annoying and prevents coda.h from simply being copied between userspace-
  and kernel-source.

I don't know; I'd guess that's just historical drift.

  - should all integer variables in coda.h be declared using (u_)intxx_t or
  not?

Probably all variables in the user/kernel interface should be fixed
width; I think this is needed for 32-bit venii to talk to 64-bit
kernels (presumably this is what you did).

  - IMHO there is no way to make my changes in the
  kernel/userspace-interface compatible to current clients compiled against
  the current (cvs) coda.h - what is the "right" way to solve this problem?

Hard to say without seeing what you did.  Was it necessary to change
sizes to be other than what the old code on i386 ends up at?

-- 
        Greg Troxel <gdt_at_ir.bbn.com>