Filing system again...

I've had a few replies about my intended implementation of a Minix filing
system server for VSTa. I should probably just clear up here that I don't
intend in any way for this to become a "real" filing system. As far as I
can see, Andy's vstafs is going to be far better suited to general VSTa
work, and I don't really intend to spend a lot of time performance-tuning
this server.

The main reasons for doing it, as I said before, were to learn how it's
done, and because I have ready references in the Linux source and the Minix
book. One other reason to do it is that it will allow me to transfer data
between Linux and VSTa with ease, without having to munge names as with
the FAT based filing systems. So, essentially, the focus is on compatibility
rather than performance (although I'll try and make the performance
adequate).

Ok, I now have another important implementation question. Where should the
disk caching/buffering be done? From a novice's standpoint, it would seem
logical to have a "cache server" sitting between the physical disk servers
and the filing systems, but I suspect that introducing another context
switch on I/O would not be a very smart idea. So, that leaves the filing
system servers or the disk device servers. Which is it to be? I haven't
looked closely enough to be sure yet, but I suspect that currently the
filing systems all maintain their own buffering information. Is this
correct? Is this the right way to go? Am I way off track?

Thanks for the replies people, it's comforting to know that there are
plenty of people out there who know more than I do about this <grin>.

-- 
pat -- empty space is wasted space.