I still believe that collaborative writing could work.
Oh sure, it could work maybe. But is it the most efficient way in
practice? I doubt it, since it forces people that are active in the document department (and those are scarce) to overly spend type on organizing and cleaning up the mess of all kinds of wellmeaning but inexperienced authors (and then I'm kind enough to forget actual abuse).
And it could work the same way as open source projects (with no/limited write access) are working. The same approach - reviews and editorials.
A very select group has write access on such systems. That is actually what is the killer, and the problem is there that the benefits of doing it online start to outweigh even the potential benefits.
There have been discussions before, the best way would have some online authoring system where users can suggest or annotate things, and that is fairly abstract so that it is somewhat machine processable. (for other format generation, mutation information for translators etc)
One could put a versioning system under the CMS/wiki system even.
But I think a plain mediawiki is simply too simple for that, and I rather have people write docs now then start a very long trajectory on developing tools. Trouble is that too many people want to write tools to maintain docs, and too few want to write docs ;-)
An edit could be corrected toward the documentation rules (which wiki-sites impose, this way or another).
True, but all this is overhead. And there have been experiments before (importing docs into a wiki with an annotation system for the next version), but the examples are underwelming.
As soon as it takes a bit of effort, and people can't edit and see it directly online, submissions drop.
Don't take such effort, and you are forever mired with heaps of very low quality edits, and the effort to actually make something out of them is larger than the value of the input itself.
The point of world writeability is to gather the information (views / ideas). Eventually someone would have to organize it.
That was the whole idea of the annotated docs too, yes. And admitted that experiment was not ideal either (there were no fixed points to link the imported docs to the next version of the imported docs).
An annotation system would work, and might be a fun job for sb interested in these kind of websystems. But all people only want to replace all existing processes with a wikipedia approach, and that is IMHO madness.
You example illustrates my point perfectly. What you see is what is there, and note that is an import of an external document into the wiki. And a bad import even.
That content is not created in the wiki (and it is the only content with an order, which says enough in itself). There is no way to compare mutations after import with the original document, no two way street. Only import and then let the forces of chaos loose on it.