Well, I've been thinking, and I think I do have an elegant, if not
likely, solution to the Linux desktop problem, but it can (I'm not
saying "will") work only if I'm right about what the most serious problem is.

The problem: despite some serious issues -- file format
exchange problems (which is being worked on, and being improved), lack
of Open Source implementations of proprietary protocols -- the main
problem with Linux being adopted on many ordinary users' desktops (for
many values of "ordinary users") is the lack of good, solid,
frequently updated user-level documentation, rather than lack of
applications, or of widespread hardware support.  

A good documentation set could be developed fairly quickly, if enough
money was thrown at it, and maintained well . . . again, if
enough money was thrown at it.  (My ballpark estimate is six months
and ten million dollars for the first pass, but I think I'm being 
conservative.)  Documentation for every flavor of Linux, of course, is a 
problem of a whole additional order of magnitude, but given the pre-eminence 
of either Mandrake, Red Hat, or SUSE, were there to be created and maintained 
a freely available documentation set for any one of the three, or of a new 
distro, that would serve many users of many distros quite well,
particularly since, if there was such a thing available, it would have a 
great impact on how future distros chose to configure themselves.  

Let me take one example, just before moving on -- where do binary
files, installed by the systems administrator, go on the / structure?
That influences some other decisions -- like, say, the path structure
for a new user -- and there's no technical reason that I'm aware of to
prefer one over the other.  /opt, /usr/bin, /usr/local/bin all seem to
be used by a fair number of folks, and all of those work just fine.
(And, in fact, Mandrake, which I'm using, uses both /usr/bin and
/usr/local/bin, pretty much promiscuously.)  All sorts of basic configuration 
issues flow from that kind of decision, which is why the KDE rpms for 
Mandrake are different than for Red Hat, etc.  

So: what large entity is likely to adopt some reasonably non-limited
Linux distro, and is going to have to produce such a documentation set
anyway?  

See http://www.nsa.gov/selinux/index.html .  

Right now, it's all in an early stage, and the folks in the project
seem, to my untrained eye, to be merely working on kernel and security
issues, rather than making a full-blown SELinux distribution available
to both internal and external "customers", but they do seem to have
institutionalized Eric Raymond's many-eyes notion, at least in terms
of this level of stuff, and all (yes, one I'm aware of the irony in
the "all") that needs to be done is for them to move forward from what
they've already done, with a view to making an entire, full-blown
distro, complete with documentation, available.

And note what they've got in their "tasks remaining to be done" list.