There have been long discussions about having a separate "Sifter" project which publishes selected revisions of existing Wikipedia articles that are believed to be accurate and complete. Such a Sifter project would exist alongside Wikipedia under a different name. Similarly, I have proposed a system for certifying articles within Wikipedia.
Both may not be necessary.
The German and the English Wikipedia currently use a quite clever process for selecting the so-called "Brilliant Prose" articles, the best of Wikipedia, so to speak. Articles are first added to a "Brilliant Prose Candidates" page, and if there are no objections within a week, they are added to the Brilliant Prose directory. If there are objections, they have to be resolved in some way.
This alone is already a kind of certification process, but it lacks one component that the Sifter project provides, namely, the establishment of trust by only linking to "safe" revisions of an article. This could be integrated into the Brilliant Prose process relatively easily.
In the article footer (where the license stuff is), there could be a "permalink" to the current revision of the article, which would simply be a link with a timestamp like in the article history. When an article is added to BP, this permalink would be used instead of a normal wiki-link. Furthermore, the BP page itself would be protected, and only sysops would actually add or remove articles from the BP candidates list to the BP page. Similarly to the "Votes for Deletion" page, sysops would simply carry out the requests of the community.
I would personally prefer if a process was in place that if a consensus cannot be reached within a timeframe, the page is added to a list of "Current negotiations", where again, for a period of 7 days, people would be invited to suggest compromises and if that *also* fails, a vote is held on the matter. This is to avoid problems like on the VfD page, where sysops are given quite a lot of room for interpretation if a "consensus" has been reached, and pages often linger without a decision for days or even weeks.
The last component that might be necessary to make this work is an associated WikiProject to organize the reviewing process. This is simply a matter of organization.
The advantages of this approach vs. a separate Sifter project: * no separate brand to the Wikipedia brand, no separate community * feedback from all Wikipedians, not just those specializing in the discipline in question -- besides being complete and accurate, articles also must be reasonably well written and easy to understand * establishes trust in Wikipedia * simple, easy to use and completely open * requires only one change to the software (permalinks), which is useful anyway for external authors trying to provide a permanent reference to the revision of the Wikipedia article they cite * Does not encourage the establishment of any POV in the selected revisions: In a Sifter project, people might just make some last minute changes and then put the revision that contains these changes on the separate site, knowing full well that the changes won't survive on the Wikipedia, whereas in this model, changes would have to survive the Wikipedia consensus process, so it works with existing NPOV guidelines
One possible disadvantage I see is that it might be harder to "launch" this project -- when there's a new separate project there's always the associated excitement, whereas a new WikiProject might not arouse the same level of interest. On the other hand, if we get too much interest, the candidate page might get too long, and we would have to split it up into different categories. Both are not unsolvable problems.
What do you think? If we do this, I think we should basically put every brilliant prose article that hasn't gone through this process in the new queue, just in case some of them might not be as brilliant (anymore) as the person who originally added them thought.
Regards,
Erik