initdiskless/rc.diskless1

Hi,
as usual at this time of the year (starting heavy diskless usage in
the labs) i am revising the diskless startup files.

As those interested know, directories in the client's root
can be filled from a cpio archive, or missing that, by copying
one of the server's directory. This has been changed  in rev 1.17
to always use both (the directory first, the cpio archive after).
I would like to rever this back to the old method, with the following
motivation:

  + i introduced the cpio archive hack because in some cases (e.g.
    paths with large delays), the separate NFS transactions for each
    file was extremely time consuming.
  + (less important) having files coming from two places makes it
    harder to figure out where a given piece of the configuration
    comes from.

Any objection ?

Also, I have one (rather trivial) patch that plan to commit, which
lets you specify extra paths in /conf/ to fetch the configuration
using the T134 bootp tag (available as kern.bootp_cookie).
The tag could e.g. contain a name for client classes, boxes with
different hardware, etc., making the configuration a lot more
flexible. E.g.

	T134="lab3 nvidia config6"

objections to this one too ?

	cheers
	luigi