>>I think your only way out is to manually pick the best pieces - unless you can code your definition of "best" into a routine which would copy the files into \my\files\TheBestOf. ScctextX.prg and a good diff tool may help speed the manual process.
>
>I was hoping source control would provide me some better method to manage the multiples and determine the best. I'll have to look into scctextX.
Depends on how you use it. From the two or three SC clients I've used so far, they all organize your stuff into one folder structure per project - which is what you probably already have, except there's the rest of the millions of files on your disk which don't obscure the view. Source control keeps only what you put into it.
SccTextX will produce text versions of your dbf-like files (i.e. classlibs and forms) with class names and methods sorted alphabetically, not in the native "last edited one on the bottom" order, so any diff application won't report hundreds of false positives just because of shuffled methods. There still will be some false positives, where the binaries left by embedded COM objects change with no apparent reason, but you'd have to ignore that manually.
Still you can only compare two files at a time - folderX\mylib.vcx with folderZ\mylib.vcx (.vca, the text version, that is)... which will drive you nuts and bolts within a few hours; I assume you may have hundreds of files, times an average of 6-7 copies of each in 2-3 versions. Just the combinatory buzz is enormous.
Maybe a better approach (i.e. what I'd do :) would be to take one project that you like for any reason (which you don't explain even to yourself) and start from it. Put that in the source control. Then as you go, you find a piece you don't like, then go look for a better version thereof in those uncontrolled projects, to replace tit with... and thus gradually build TheBestOf version.
Of course, if you have an algorithm which will just know which file is the best, go for it... (g, d & rvvf)