Re: Wishlist: credential file update/access collision avoidance?

On Mon, 2011-10-03 at 15:16 -0500, Nico Williams wrote:
> On Mon, Oct 3, 2011 at 3:03 PM, Jeffrey Hutzelman <[email protected]> wrote:
> > On Sun, 2011-10-02 at 17:57 -0500, Nico Williams wrote:
> >> Oh, and on Unix systems, ccaches should only ever be stored in tmpfs
> >> filesystems, so, yes, you can expect POSIX semantics.
> >
> > Um, reality check?
> > I agree that storing file ccaches in such a filesystem is a good idea.
> > However, not all in-memory filesystems share the properties of Solaris
> > tmpfs.  "POSIX semantics" covers an awful lot of ground, and I'll bet
> > that most filesystems, in-memory or not, are incompatible in some way.
> 
> What we mean here (what I take to have been meant) by POSIX semantics
> here is not necessarily strict adherence to POSIX, but things like
> "unlink() of open files is allowed" and "rename() is atomic and
> unlink()s whatever file might have existed with the target name" --
> that sort of thing.

That was not at all clear to me from context.

Certainly I think it is reasonable to assume that a filesystem used
for /tmp will have those particular properties.

-- Jeff