So, if you look at some of the factors that he claims "dirty" the data, one is that a lot of management/refactoring of code is done by Sun employees, and another is that a lot of outside contributors find it hard to get commit access and just leave patches in the bugtracker, which Sun employees pick up.

He claims this means that the graph — which shows Sun responsible for the vast majority of all the commits — is inaccurate.

I claim that if the project isn't getting external contributors, or not giving them the ability to get commit access, and it's basically being run by Sun internally, and he has to mention that people leave patches in the bugtracker, that itself is a sign that the project is profoundly sick, more so than any charts can so.

As further evidence that Sun isn't playing well with external contributors and the community, read the tale of the non-upstreamed Calc solver, which Meeks linked to.

This is not true at all. Sure, you can type stuff in, mark some stuff bold, spell check it, and print it out -- but there's no need for an office suite to do that, and if that's all you intend to do don't call yourself an office suite.

Here's something I ran into yesterday. There's a "Compare Documents" feature under the Edit menu. It doesn't compare the contents of tables. The bug reporting this was opened in July 2003, and nobody has seemed to care yet. In 2007, someone had a patch, which was committed and not added to the next release's codeline because "I don't think that this issue fulfills the criteria for 2.3.1". This may it was retargeted for 3.1 and rejected in November because There are too many open questions to finish in 3.1." People complained again in 2004 and 2008; I don't think you can say in good faith that "no one cares enough".

It occurs to me that your exact phrasing was "no one cares enough to add it", which is completely right. Nobody cares enough to develop OpenOffice.org to where it should be.

If you ask what more, are they not done, then I'll ask the same thing about the Linux kernel -- isn't it done? What benefit is there to running the latest 2.6.28 or whatever instead of 2.4, which worked fine for everyone a few years ago? But yet who in their right mind would (all other things being equal) set up a new system with 2.4 instead of some kernel released this year? And you'd laugh if I suggested the Linux 1.x tree, but that can open and close programs and files just as well as any other OS, can't it?

No fine-tuning user access, no support for external authentication

MediaWiki supports both of these. The first is the userrights array, which you can configure in LocalSettings.php, and groups configured in the usergroup table. The second is by subclassing AuthPlugin.php and telling it exactly what you want — autocreate local users? create remote users? fetch a password db or authenticate externally? replace the login and logout pages?

Most decent external authentication providers set REMOTE_USER, and meta has a plugin for autologin with REMOTE_USER.

not easy to add new markup

I haven't had a reason to try this, so I don't know. But you can make pseudomarkup really easily via the template feature, e.g., map {{{serif|foo}}} to <font face="my favorite serif font">foo</font>, or whatever. In fact, templates are how Wikipedia does their infoboxes (the tables on the right side of articles about prominent public figures or businesses or whatever).