"revision" considered confusing (was: Outdated Wiki Page?)

[John Yates]

On Tue 2007-04-24 Aaron Bentley defending the term revision wrote:

> It also doesn't apply to VCSes that don't store snapshots, like
> Darcs or, depending on your POV, Gnu Arch.

I know nothing about Gnu Arch though I wonder whether Martin's
delta/snapshot equivalence argument might not apply.  While you
as an implementor may care about underlying representations when
I explain vc to non-programmers I find the image of a succession
of snapshots the clearest expositional model.

As for Darcs, I know that claiming that record creates a new
revision is quite inaccurate except in some Humpty Dumpty sense
of appropriating a term and ascribing to it unfamiliar meaning.
Darcs record creates a function or patch.  Darcs documentation
-- especially the discussion of patch theory -- is quite clear
that the record operation creates only a patch, not a snapshot.
A patch is a true first class object.  It is tied to neither an
input nor an output tree signature.  This fact underlies Darcs
exceptional support for cherry-picking.

Trying to explain Darcs as recording a succession of revisions
or snapshots is a recipe for future confusion when delving into
more advance capabilities.

/john