RFC: make the experimental NFS subsystem the default one

[Rick Macklem]

> 
> My main interest is interoperability with systems other than FreeBSD
> such as Solaris, Linux, and Apple's OS X.

I have Leopard and Solaris10 and do interop testing with them. I have
done some with Linux (and had a couple hundred Linux desktops mounting
the server a few years ago), but since there are sooo many Linux variants,
it only indicates some have worked at some point. (The server also saw
Snow Leopard clients for a while, but my Mac is an old G4 Powerbook so
its Leopard for me:-)

> Client throughput of
> FreeBSD's default NFS is low. I would be hoping to get more
> throughput. Also, modern NFS uses TCP so I would expect to use that.

Hmm. At least for read rate on a large file, others see wire speed at 1Gbps
and I think some get close to wire speed on 10Gbps. (I see wire speed for 100Mbps,
but that doesn't say much:-)

Was it read rate you were referring to or something else?

I would certainly consider it a bug if people see poorer perf. for the new
client than the old on the same setup. (I think the difference in read rate
observed some time ago is now fixed in head. The experimental client was using
a default readahead of 0 instead of 1.)

> I am also using the automounter and hope that the updated NFS works at
> least as well (or better) than the default one.
> 
Righto, I need to test that.

rick