Charles Duffy <address@hidden> writes:
Losing dumb server support seems a bit much to pay for such a
generally unimportant feature.
Storing binary files more compactly seems pretty unimportant, I agree
(well, it seems unimportant to me).
Storing text files in such a way that they can be retrieved
efficiently seems more valuable---so valuable that we're willing to
store multiple revisions in plain text.
I think it's at least possible that per-file things (such as "cvs
annotate" and the like) are more commonly used than things like
star-merge, and so a version control system implementation that
optimised those (and made things like star-merge less efficient) might
be a better fit for what people use version control systems for.
You wouldn't need to lose dumb server support---there'd be nothing
stopping an implementation also producing and reading Arch archives.
Indeed, this xdelta storage of revisions could simply be a more
compact revision library implementation---that would surely be
uncontroversial.