On 19/07/07, Ray Saintonge saintonge@telus.net wrote:
Talk page copies of a history are useless because the links aren't there. They let us know who edited, but give no details about what edits they made.. A history merge may not accomplish that either. Merge and redirect would not destroy the history. Could a special line be devised on a history page to register a merger, and allow one to trace bacvk into it?
It really doesn't seem essential for the GFDL, now I think about it. We're tracing the authors of *a document*, not of individual words and sentences. The article's a collective work, a work by many people, not an aggregation of a hundred or a thousand tiny seperate works. If we had no history section at all, just a little script somewhere that added your name to a list of "this page's authors" whenever you edited an article, we'd be fine.
For editorial purposes, being able to see the details of the history is really handy. For license purposes? It's icing on the cake.
We have a list of contributors to each article - the list on the history file. We dump that onto the talk page as a plain text list. If we can look down that list of names and say "these sixteen names are there" we can easily turn it into "the document, as of this moment, is a collaborative work by these sixteen people", and we're done. Your authors are named; attribution is done and dusted.
All we need to be able to say is "the following people worked on this document". There are some false positives - people whose edits were quickly reverted and made no contribution to the article, say - but they show up in the history anyway, and I guess you could argue that they were in some conceptual way fellow workers on the document!
...
There are, of course, downsides to dumping the history; it removes the ability to do in-depth analysis to figure out who the "real authors" were. But this is less of a loss than it seems, because this is an astonishingly impractical sort of analysis to do - you can winnow out the obvious not-involved-in-the-work cases, the vandalism-revert pairs, but after that it gets very hard to computerise and awfully easy to miss out real, significant, contributors - lots of false negatives. Once there's been any kind of complex back-and-forth, simply running an automated tool for "biggest contributors" is going to get some pretty inaccurate results in any meaningful sense.