From an email with the author:
"On Unix, I used RCS fairly happily. It was too verbose, but I fixed
that. RCS, like most version control systems, also believed that the
RCS archive or repository was authoritative and I don't accept that, I
think that the checked-out source is authoritative. If you are
grepping source, you really don't want to grep the archive files, and
you shouldn't have to check out source just to compile. I view the
archive as just history. So I added options to do things my way.
For my purposes, the Plan 9 dump, cp, diff and idiff work well. I
don't miss a formal version control system.
Git is a horrific botch of a version control system: wasteful, slow,
stupid, and unpredictable. When I check in source file(s), I want a
snapshot taken always, and I don't want to be told, no, you can't do
that, you have to rebase. The fundamental problem is that companies
want to believe that multiple programmers can edit the same program at
the same time. That's a great way to introduce bugs and break things.
When editing a program, you want to have a stable view of it, so you
know what you can rely on, you don't want it to be shifting underfoot.
The idea that you can apply an arbitrary set of diffs, or subset, to
some version of that program, and get a meaningful and correct program
out is absurdly optimistic; I wouldn't trust it."
