I've been thinking about what KDE is. Or what it means to be a "desktop environment" at all. KDE has many faces (and we're not just talking about kde:KJanusWidget 
. The windowing manager (and associated KDE Panel) is the public face, but the infrastructure in
kdelibs that makes the key applications (KMail, Konqueror, konsole, kdevelop, and lots more) not just possible, but consistent and reliable. The toolset in kdevelop, and the developer community are key parts of KDE too.
But what is a client server architecture without servers? Maybe we should have a K Server Environment which provides excellent support for the K Desktop Environment. I'm thinking of something that is based around the Kolab project, and extends it out using some of the other
key server projects (eg Samba). The development environment would need to include a no-X version of Qt and some of the kdelibs, plus perhaps some of the networking code. This would be a good time to split kdelibs into a couple of parts. I guess the model I have in mind is
something like the E-Smith (now Mitel SME server, recently orphaned) distro, which provided a uniform approach; although implemented with more of the LDAP concepts of Kolab, and without the half-firewall content (since the firewall should be dedicated IMHO - the K Server would be hardened to live in the DMZ if required).
The general idea would be that you have a workgroup server supporting a set of machines with KDE. The underlying operating system would be inherently irrelevant - this server could be on Linux, BSD, Solaris, whatever; although I'm drawn to Linux or BSD as for the auto-install idea. The installation would be mostly automatic - you put the CD or DVD in, you confirm you really do want to trash the current contents of the machine, and it installs. At the end, a short wizard walks you through the essentials of post-installation configuration. There are a range of meta-distributions that can already support this, although some tweaking may still be required.
The advantage here is that we can tune the server side to optimally support whatever the desktop would ideally use. Sure, other servers would work, perhaps even as well, but the K Server Environment (KSE? KDE-Server? KServ?) would be a consistent, tested platform.
Syndicate Blogs
More new standards?
Correct me if I'm wrong, but would this not open the door to lots more potential incompatability between KDE and other competing platforms? I know it would be easier from KDE's point of view if some distributors put together servers tailored for KDE, but in the long term it seems wiser to look at how desktop environments interact with server environments in general, and perhaps to then work through something like freedesktop.org to improve this interaction so that all WMs/DEs can benefit, so that all related technologies can make themselves compatable, and most importantly so that users don't become forced to use a particular application or solution because others are unable (as opposed to unwilling) to take advantage of a particular feature.
A general review of interaction, OTOH, which I think you'd support (?), could be quite a good idea. It'd be especially good to get distributors in on this, as they all provide their own crazy tools for things like managing samba and workgroups. Then distributors, and particularly meta-distributors, could create systems with far more seamless interaction without putting a huge new burden on KDE developers and without shutting out all the users and developers who don't want to exclusively use KDE Project products.
K (server) distro
I think what you are talking about here is a K-server-distro. (i think the easy install you discribe could not be done without providing a hole disto) And the idea is very nice... But it should have a solid base IMHO; so eiter (1) a company backing it up, or make it (2) an extension to the Debian (or Gentoo) Project and provide strong community support.
When I think of implementing it in my company (6 desktops) I would prefer the 1st situation, and have some nice support contract.
Anyway if Kolab gains momentum the server based distros will add it in; and will provide have support for it; and we can hopefully control it all (kolab, samba3, other-autentification-stuff, remote-home-dir, you-name-it) by LDAP, by default. So we only have to add/remove users once!
Ok, that's it... See y'all in the future! I'm going into HyperSleep(tm) till this happens -- Bye bye!