On Tue, Aug 05, 2014 at 07:16:57AM -0400, Greg Troxel wrote:
[Coda r/w performance]
> > ... is about processes efficiently accessing the _local_ (cached) files.
 ...
> > (I recall ZFS on Linux has remarkably worse performance via FUSE compared
> > to natively)
> 
> That's a fair point.   What I really mean is that the speed of FUSE is
> high enough that it doesn't seem likely to be an issue.

For new platforms, yes. I guess it is what you mean.

(when there already is a Coda-specific kernel module, FUSE would
noticeably limit the efficiency/throughput of the r/w operations -
as they are locally bound)

> It's also possible to add in some sort of container file interface to
> FUSE, eventually, which is similar to what coda does, but gets it moved
> to a different layer where hopefully more people/code will use it and
> care.  Coda just has too few people...

A "modified FUSE" is conceptually the same as a "Coda-specific kernel
interface", unless you convince the world that the wannabe new feature
is useful for "everyone" or at least for "many", besides Coda :(

Rune