Re: cvs rtag -r BRANCH -D date

[Ken . Olstad]

"Derek R. Price" <address@hidden> wrote:
> if we go with '@', I'm going to say address@hidden is probably the more
> generally intuitive way to select a point (date) on a branch due to
> the meaning already assigned to '@' in other (non-cvs) contexts.

Ah, interesting.  Perhaps you're thinking of email addresses
address@hidden, with USER being subordinate to DOMAIN on the system, and
likewise DATE is subordinate to BRANCH, hence address@hidden  I can see
that, but doesn't address@hidden scan only because of the identification
of USER with a person, and DOMAIN with a place?  address@hidden thus
means "a person at a place", which is a case of expressing
primary-then-secondary, not vice versa as address@hidden seems at first
glance to suggest.

With BRANCH and DATE, I think primary-then-secondary is also the way
do go.  This is why I think address@hidden is more intuitive.

1. address@hidden is English-like, with "@" pronounced "at" and meaning
"as of", so address@hidden reads like: "the state of BRANCH (a project,
task, release, etc.) at DATE."  How do you read address@hidden

2. The branch feels to me like a higher organizing principle than the
point in time.  Each branch is a copy of the universe, and has its own
timeline.

         branch > year > month > day > hour > minute > second

BRANCH is a coarse specification.  YEAR refines it, MONTH refines it
further, etc.

3. A branch has concrete and useful meaning without a point in time.
The converse is less clear to me.  (This item is a lot like #2.)  I'm
quite sure I've never set up a sandbox with a single sticky date and
multiple branches, i.e. some files from one branch and others from
another, but all with the same date.

4. There is already the BRANCH:DATE precedent, as with cvs update -j.

Ken