[OpenAFS-devel] Re: Fuse and testsuite

[Simon Wilkinson]

On 26 Sep 2011, at 00:02, Andrew Deason <adeason@sinenomine.net> wrote:
>=20
> You can also go the route that Derrick suggested where we use the
> existing rmtsys mechanism (which is usually used to forward pioctls to
> another machine).

Strategically, I think it would be nice not to add a dependency on the rmtsy=
s code. At the moment, the only user of that functionality is the NFS transl=
ator, which is becoming unsupportable on today's operating systems. Having s=
upport for rmtsys on the client side adds a lot of complexity to our code. I=
t would be nice to consider a world in which we can build OpenAFS without it=
.
>=20

> Also, to get authentication working, just get the pioctls working. aklog
> just sets credentials via a pioctl, so once it can work with them, it
> shouldn't need anything else special. I think libuafs (or another
> library) really needs an API that allows setting credentials more easily
> (so you don't need to issue the pioctl manually in whatever
> libuafs-using application you're using), but for the purposes of the
> fuse client it's not necessary.

One thing that I am currently working on is turning aklog and friends into a=
 library. With this, and the new token code, it should be possible for the l=
ibuafs API to simply expose a "set a token" function.

>=20

S.=