Damn it. Firefox just crashed, which is very rare these days, still though. Argh. Why does a browser always crash at the end of writing a long post? sob.
It isn't that a Linux desktop cannot be built, but that a business model to support a linux desktop has not been identified (assuming that one exists). Redhat splintering off Fedora is a good example of the difficulty involved in creating a Linux desktop. Several others have also given up or failed to develop successful businesses around the desktop - Caldera and Corel among others. And before somone says it, I know that Debian and other volunteer based distros have been around for a long time and offer a good desktop, however, they haven't made much, if any, inroads into MS's marketshare. I don't see why this will change going forward. Linspire (or whatever they're called now) and Sun have interesting licensing models, Novell might surprise us too.
I think that MS decision to neuter Longhorn down to an XP upgrade will most likely put Longhorn in the back of the line. WinFS and Avalon - which are evolutionary MS upgrades - are being cut from the Longhorn features.
AFAIK, Avalon is still a go. This has certainly been an intersting time period for MS. IMO, they're screwing up at an alarming rate with the sp2 rollout and now having to cut back on Longhorn, which was suppose to be this new great thing and is sort of turning into just barely a thing at all. Anyways, the business model is very important and hard to break. People are resistant to change.
2) Project Evil: yes, this is a FreeBSD project, but it has to get ported to Linux. Why? It will increase the number of devices that Linux is able to run.
Project Evil only works for wireless nics, and even then not all of them, eg. usb nics don't work, so we're back to convincing vendors to open their hardware or support Linux. You are right though. Any os that hopes to displace Windows will need excellent device support since the world has grown accustomed to having everything just plug in and work.
3) I can't emphasize the need for a standardized desktop API. Qt, GTK, GTK2, etc.. There are far too many API's, we need a centralized solution with backwards compatibility.
Yeah, well.... you know.
I was thinking about something similiar to this earlier today. With .NET and Java, the applications can be removed from both the OS and desktop (window manager), so it may be possible to have something like.....
java >> looking glass >> solaris
.net >> window maker >> os x (the mac crowd: "pfft! Aqua!")
There are some pretty huge issues to overcome, but I think it is an interesting idea. When I have more time....
4) Longhorn-like features. Though this might seem stupid, but yes, GNOME has in works a new filesystem, etc. I think that if they all collectively worked on features that Longhorn couldn't deliver, I can see Desktop Linux as successful.
Two quick points......
First, there are some intersting cutting-edge open source projects like Looking Glass. However, many of these projects are never adopted because they sprung forth from a company, which 50% of the OSS community isn't going to like for a million lame reaons, and doesn't use license xyz or language abc, or some other damn thing gets in the way, and eventually anyone who asks why is greeted with....
OMG WTF, RU EV1L? We would never include that inferior shit in our distro.
Sometimes OSS biggest enemy is OSS.
Second, I think that whoever wrote the Halloween document, MS or ESR, got it partially right with 'OSS follows in the tail lights'. OSS is very suitable for older, well established, things... compilers, good - games, bad. There are several reasons and quite a few exceptions, but I think this is a pretty good rule of thumb.
Prior to Linux, you could even argue that OSS doesn't do operating systems very well (bare with me). After all, Linux is a bit of a fluke. Most other OSS os'es have mostly failed, and the 3BSD's were based off of the AT&T source code, and while they are entirely OSS now, the design is still based and influenced by the old BSD. Who knows, but would FreeBSD have been successful without the AT&T template?
I'll add something on MS and the software lifecycle later.