Kindness!? From me!? Just kidding...
That's why I'm working with the existing openafs branch. When I added
OpenAFS support, I did not rip out NFSv3. In fact, I added a lot of code
like this:
given ( $celltype ) {
when ( q{afs} ) {
# new openefs code
}
when ( q{nfs3} ) {
# existing "legacy" nfs3 code
}
when ( q{nfs4} ) {
# nothing yet....
}
}
So a lot of the work to identify where each "celltype" is handled, according
to the underlying distributed filesystem protocol, has already been done.
I am assuming, and I think it's reasonable, that NFSv3 will always use
sec=sys, and that NFSv4 will always use sec=krb5 (or perhaps krb5i or krb5p
even). I may change that as I get this working though. Then again, I
don't intend to use NFSv4 without Kerberos, so unless someone has a real
world need for insecure NFSv4 support, I don't know why we would want to put
the effort into supporting it. We'll see....
Practically speaking there no reason to drop the relatively insecure NFSv3
code at all, and eventually, I think we can even extend the test suite to
run all three modes in one test run (that will take hours, but...). The
test.efs set up already has both AFS cells, and NFSv4 exports in place, and
we have *never* run the test suite with real NFSv3 mounts anyway (they have
always been faked out with symlinks to the local filesystem on the two test
efsservers).
One last, somewhat cynical point: what work on the existing NFSv3 code are
you refering to? There have been ZERO commits to efs-core since I left,
and ZERO discussion on the mailing list, so as far as I can see, there
hasn't been any.
On Wed, Jul 13, 2011 at 8:52 PM, Steven Jenkins <[email protected]>wrote:
> On Wed, Jul 13, 2011 at 5:14 PM, Phillip Moore
> <[email protected]> wrote:
> > I just committed a set of changes to efs-utils, specifically
> > efs_virtual_machine, that adds the necessary configuration changes to
> enable
> > the use of NFSv4 in the test.efs setup. Over the next couple of weeks,
> I
> > will be working on the hard part: changes to efs-core....
>
> I'm excited that you're working on this again.
>
> However, I have a 'plea for kindness': please either do this in topic
> branches, or make it multi-platform (ie, able to set up either NFS v3
> or v4). I strongly prefer the latter, but I realize it's more work
> (and the benefit doesn't come to you in the short term), but
> especially once you get into efs-core, the changes for NFS v4 will
> make it much harder to merge work being done on the NFSv3 product.
>
> If you're already planning on making this dual-platform, "thank you!".
>
> Also, I'll make you a deal: if you'll do the work in topic branches, I
> will volunteer to work on merging into master to make cross-platform.
>
> Thanks,
> Steven
> _______________________________________________
> EFS-dev mailing list
> [email protected]
> http://mailman.openefs.org/mailman/listinfo/efs-dev
>
_______________________________________________
EFS-dev mailing list
[email protected]
http://mailman.openefs.org/mailman/listinfo/efs-dev
