RE: Official sources vs. RCVS

[Cameron, Steve]

Karl Fogel wrote:
  [...] 
> Seems useful; if understand it correctly, it can alleviate the need
> for branch-point tags, right?
> 
        [smc]  Yes, that's what the ".origin" part does

        And, the ".trunk" part
        can do essentially what "cvs update -A" does, except
        it doesn't fry your "-kb" and "-kk"  sticky options, 
        so you can switch between the trunk and a branch 
        easily, and allows you to do "cvs diffs"  or other 
        operations vs.  the trunk with out relying on using 
        numeric revision tricks that may not always work 
        if your repository  was made by directly importing 
        RCS *,v files, or if  somebody has unwittingly done 
        "cvs commit -r 2.0"

        One questionable aspect of the patch is that it
        allows the combination ".trunk.origin", whcih will 
        return the oldest revisions on the trunk.  I call it 
        questionable because as time passes and files 
        are "cvs added" to the trunk, 
        the list of files that ".trunk.origin"
        refers to will grow, unlike "branchtag.origin", if I'm
        remembering this right.  I think this is because
        when files are cvs added to a branch, a dead
        revision is added first, but when adding a new
        file onto  the trunk, no such dead revision is 
        added first.  In any case, the combination of
        ".trunk.origin" as a single tag probably isn't too 
        useful anyway.  It should either return what it 
        does now, (a list that grows over time), or, the
        empty set, which would be fairly useless, or,
        maybe it should just be forbidden, as senseless.
        I implemented what seemed to the least senseless
        thing without forbidding it altogether.


        -- steve