Re: Vnode scope

[Elad Efrat <elad%NetBSD.org@localhost>]

Ken Hornstein wrote:
> > Limitations:
> >  - It's confusing that VOP_ACCESS() doesn't check access control, but
> > we can document that...
> >  - ACLs are limited to what our normalized representation can
> > represent, but that's not a problem for now I guess
> >  - Lots of changes to replace VOP_ACCESS() to vnode_authorize()
> >
> > Any thoughts on the above? anything I missed? do we want to do it that
> > way? anything we should take into consideration?
>
> So, some things come to mind.
>
> First off, I don't see how this could work with, say, an AFS client.  People
> will point out that we don't have an AFS client for a modern version of
> NetBSD; that is true, but it's not an insurmountable problem; it just needs
> someone to dedicate time to porting it (I wish I had the time; sadly, I
> do not).

First, why wouldn't it work with an AFS client? (I don't know AFS)

Second, I don't think we should hold back on features, especially ones
that have the potential to work right away with all file-systems we have
have *and* add ACLs to them -- all using kernel code that's already in
place -- because of one file-system that isn't even implemented.

I've heard "we're getting ACLs through extattrs very soon now" for
literally years, and given they're supported in FreeBSD and we have part
of the code in NetBSD, I'm going to be *very* skeptical about anything
before it makes it into the tree.

> I don't think the normalized ACLs that NetBSD has are rich enough to support
> all of the possibilies that are out there (especially in the case of network
> filesystems).  Sure, it might work for our current filesystems, but what
> happens when we want to add something new?  What, exactly, are we supposed
> to do?

Read my other replies stating my solutions for not binding ourselves to
a normalized format. That said, I think what you're suggesting we should
be worried about has a very, very low probability of affecting us.

NetBSD has supported only traditional Unix permissions for 16 years.
Even though our kernel has some initial code (and just that) support for
extattrs (and ACLs) in just one file-system, nobody ever really cared
about implementing it. Trying to claim that a normalized form, that will
work with, what, ~10 file-systems we have now, will become irrelevant
when we add a file-system with features that we did not foresee is going
to be a problem is both non-realistic and a "rich man's problem"...

Thanks,

-e.