[patch 5/6] Implement a new version of lucid spkm3 context.

[J. Bruce Fields]

On Tue, Oct 17, 2006 at 09:54:06AM +1000, Neil Brown wrote:
> On Friday October 13, trond.myklebust at fys.uio.no wrote:
> > On Fri, 2006-10-13 at 15:35 -0400, kwc at citi.umich.edu wrote:
> > > @@ -51,6 +51,7 @@
> > >   *
> > >   *	buf->length should be:
> > >   *
> > > + *      version 4
> > 
> > How is this versioning useful? There doesn't appear to be any
> > provisioning for a fallback to previous versions for kernels that don't
> > support version 4. Am I missing something?
> 
> Ping? I would like to see the answer to this one too.

I'm hoping Olga or Kevin will remind me what the deal is here.  I've
forgotten.  As best I understand it:

This code breaks the interface--the kernel will no longer work with
versions of gssd that worked with the old kernel.

I understand why that's an awful thing to do in general.  But the
versions of gssd that used that interface were experimental code only
available from our website, and probably, alas, haven't even been used
outside of citi.  So maybe we can get away with breaking compatibility.

But I definitely don't want that to happen *again*.  So I'd like to make
sure we have a way to extend the interface in the future without losing
backwards compatibility.

Adding a version number--even though it's just a placeholder for
now--seemed like an obvious way to do that.

But maybe it's dumb.  I'm open to suggestions.  Or to just saying "too
late, we have to maintain that old interface even if you think it's
obscure."

--b.