My requirements for a wiki:

1) Awesome search. Xwiki has a very good search, including in attached documents

2) Fine grained access controls. I need to be able to have a space W with pages X,Y,Z, and Z needs to be public, but nothing else.

3) I need to able to use my existing authentication systems.

4) It needs to be user friendly. XWiki has some serious usability problems for anyone who hasn't used a wiki before. Confluence is a bit nicer.

My nice to haves:

* Site-based documentation. Sometimes I have a workflow that's 80% similar, but the exact steps depends on the site (Example: ssh to this host, run these commands). I don't feel that it justifies having a separate page for each site when the content is mostly the same.

* Editable from my Linux environment, preferably via command line

* OneNote integration

I've used mediawiki, xwiki, twiki, and confluence before.

If you can provide me with most of the things above, and either the source code, or a service SLA (or both), I'd probably buy.

I've been through this cycle a few times, this time we are using MediaWiki. I don't remember what we used last time, I think it was a Drupal-based wiki.

None of them made a whole lot of impression on me either good or bad.

I agree w/the other posters that this is a pretty ambitious goal ("better wiki") - I'm like a lot of other folks in that what is out there is more than enough and we probably don't even exercise 10% of the features.

I also agree w/the other poster that said "people don't bother". Very true, stuff usually doesn't end up in wiki unless I put it there or I ask somebody else to do so. Very few people will do this habitually without a lot of strong and recurring encouragement.

Meadiawiki, pbwiki and Confluence are at the top of my list of wikis with really poor usability. I think the WYSIWYG editors are in most need of fixing (Write your own - if you can't do this you should not be building wiki software) followed by tracking changes and linking of documents together in a manner that allow you to easily track groups of interesting docs without being overwhelmed by updates after a certain point.

The last thing the world needs is more poorly done wiki software so aim high :-)

Before that, it's just a lot of effort to maintain compared to the benefit that arises from it.

EDIT: But I personally wouldn't pay your company for wiki software unless it is seriously good. There is a lot of free wiki software that is good enough. Also, for non-technical users, an internal wiki will be less useful. It only works when people are interested in editing it, and in my experience that is pretty much only technical people.

Internally, the dev teams and the IT staff use the wiki, but externally everyone uses Microsoft OneNote and there is no way for us to possibly convince them that a wiki would ever be better.

OneNote seems sort of like a wiki for the masses, and I think if you could create wiki software that is as easy for lay people to use as OneNote, you would have yourself a marketable product.

I think Ward's Smallest Federated Wiki is an interesting direction to go, and should really be taken seriously.

For one, it assumes multiple people want to edit it. People don't bother.