[ILUG] Windows 98 Replacement?

[Rick Moen]

A week ago, I wrote:

> Quoting fuzzbucket (fuzzbucket at eircom.net):
> 
> > I'm not sure if I want my system to keep ticking over if something as
> > vital as block devices start going a bit wobbly. I think if there's a
> > chance of data corruption I'd prefer my kernel to panic and sieze up :)
> > 
> > To use an entirely inappropriate car analogy (I'm gunning for a CEO
> > position) would you like your engine to keep going if you steering
> > stopped working?
> 
> The example I had in mind was the various schemes for handling crypto
> filesystems:  Filesystem integrity is checked at that level, not at the
> kernel level.
> 
> Sorry to not have examples handy.

As it turned out, what I was unsuccessfully trying to recall to mind was
the FUSE architecture (Filesystem in USErspace) patch, 
http://fuse.sourceforge.net/ .

The kernel folks, however, have indicated that they're unlikely to ever
accept it, for reasons Zack Brown has summarised here:
http://www.kerneltraffic.org/kernel-traffic/kt20050103_289.html#1

Notice the quotation from Torvalds where he explicitly rejects (yet
again!) the logic by which microkernels justify the loose integration of
such filesystems.  Torvalds feels that filesystem behaviour needs to
remain under the close supervision of the kernel team.

Which points out one difference between that and NT4's move of all 
video support code into ring zero (kernelspace):  That sets up a
situation where even a _screensaver_ can and does bring down the entire
host.