--sdtB3X0nJg68CQEu
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Content-Transfer-Encoding: quoted-printable

On Wed, Apr 21, 2004 at 08:36:31PM +0900, YAMAMOTO Takashi wrote:
> hi,
>=20
> > I definitely think the "no way to see" aspect is bad. I think there may=
=20
> > still be value in keeping the removal, but we definitely should let the=
 fs=20
> > know there was/is an entry.
>=20
> for a little better example...
>=20
> even ufs can take advantages from (non-whiteout) negative entries
> because, in that case, it doesn't need to search the entire directory
> to check whether or not the entry exists already.

True.

The one issue I'm having with this suggestion is I'm not sure how to=20
handle the case where we find a vnode (for ufs). We still want to scan the=
=20
directory for the entry, but it would be simplest to keep the vnode around=
=20
while we do it.

Take care,

Bill

--sdtB3X0nJg68CQEu
Content-Type: application/pgp-signature
Content-Disposition: inline

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.3 (NetBSD)

iD8DBQFAhqBpWz+3JHUci9cRAooQAJoD49mZu83olU3FzeLQTN9aJSdexwCeIlTa
Hkj4rDU1fHI7ydAh3tE9r7M=
=Ve4t
-----END PGP SIGNATURE-----

--sdtB3X0nJg68CQEu--