I agree (would mod up but gave up modding way back). However this is an interesting and probably reoccurring problem: extending the wealth of public net wisdom with precision data from local context (organisational or task-centric rather than geolocational).
A proxy adding local content into pages loaded from outside as suggested in Re:Solution by mcrbids would solve some of the problems you mention:
* The wikipedia content will always be out of date
* it's fetched from real sources in real time
* Changes made to wikipedia content don't get fed back into wikipedia
* this changes to risk of unintentionally publishing private information - how hilarious!
* Trying to load the wikipeida DB locally is a headache due to its shear size
* not done; could instead cover the whole of outside web with one solution.
This problem remains:
* Creates confusion as to what is and is not company information
I guess you'd be best off injecting a (user-hidable) "widget" layer that would contain all the local information needed, thus providing clear separation of local and global content. Least breakage of existing layout that way, I hope.
I assume here that we restrict our proxy to embed HTML (possibly including Javascript) into well parsing HTML pages only, so as to avoid breaking things as much as possible - inevitable to happen sometimes anyway.
Updating the contents of another window based on browsed content would require either
* a single sign-on solution to target references to correct user's desktop (seen in updaters of multiple applications views in medical solutions for instance) or
* a browser-specific local hack to study each page url and content and to fetch related information from local database based on those.
OT, adding such meshing into Google Wave would probably prove an interesting challenge :) Think of doing it Right (tm), with private additions to documents and discussions getting saved and tracked on local servers while public parts would be passed on to public servers.