Don't worry. This _will_ happen - it's pretty high up on my list,
actually. The reason it didn't make it into 2.0 was that I considered it
less important than read caching (in general I tend to think read
performance is about ten times as important as write performance), and
because it has some very nasty issues that I didn't feel comfortable
about fixing in the late 1.3.x series.
My _personal_ plans for 2.1 will be the i386 kernel mapping change (kernel
mapped at high virtual addresses, no separate user segment) and right after
that (it should be a few days of coding once I get started on the 2.1 tree)
I'll start on the NFS and networking stuff.
(In fact I hope that Olaf may beat me to it when it comes to NFS - he
knows NFS much better than I do anyway).