On Wed, 13 Jul 2005, James Buchanan wrote:

> Since we can't know the physical disk geometry on modern disks,

I don't believe this is true.

If one were willing to measure transfer rates for a variety of sectors, I suspect an exact map of the disk could be determined.

If the disk can tell you when it is re-mapping a sector, one could adjust the algorithm.

I believe this uncertainty can be completely removed.

> it doesn't 
> make a lot of sense to design a file system with knowledge of cylinder 
> boundaries or any other disk information.  A new file system should probably 
> be a simple layout of extents of size x (probably page size, and configurable 
> at file system creation time).



> Can a new file system somehow be implemented in userland first before adding 
> it into a kernel?

Yes.  Simulate it.


> --
> James

-Mike

p.s. I have for years now advocated a tighter coupling between the VM and a huge FS.  Others with less vision have tried to shoot me down.  Well, I'm still here, and we still need tight VM/FS integration.