--9amGYk9869ThD9tj
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Content-Transfer-Encoding: quoted-printable

On Wed, Mar 08, 2006 at 02:18:00PM +0900, YAMAMOTO Takashi wrote:
> > > i don't think of any usage for which inclusion of "nfs" is useful.
> >=20
> > I can! Any tool which wants to know what io stat names are available.=
=20
>=20
> for example?

I don't have one handy, but I do know of applications that take the=20
current disknames sysctl and applies a regex to it to determine as list of=
=20
drives that are candidates for certain internal operations.

> > I think a better thing to do is rename the variable to "iostatnames". T=
hen=20
> > it makes a lot of sense to include nfs in the list of all iostat-provid=
ing=20
> > sources.
>=20
> for what is it useful, in addition to statistics themselves,
> which include those names?

So that a program can know what statistics are available.

I admit I'm operating a bit in a vacuum as I haven't tried the latest=20
patch. But the difference is one of explicit listing vs. implicit listing.=
=20

Yes, we can know what stat types are available by looking at the mass of=20
stats. But I think there also is value in directly knowing what stats are=
=20
there.

Take care,

Bill

--9amGYk9869ThD9tj
Content-Type: application/pgp-signature
Content-Disposition: inline

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

iD8DBQFED0xyWz+3JHUci9cRAtvvAJ92cUxxpsYGlTZy+rhu4xoGLZRkfACfb49D
/X7UsflNDE3eQ/e2Me6HwYY=
=lBxc
-----END PGP SIGNATURE-----

--9amGYk9869ThD9tj--