Definition of branch?

[John Yates]

On Wednesday, 2008-01-23, Ian Clatworthy wrote:

> Tim Hill wrote:
> > Sorry to spam with such a simple question ...
> >
> > The User's Guide says: "A branch is an ordered set of revisions" and
> > "Branches may split apart and be merged back together". From this, I
> > take the definition of a branch to be a linear timeline of revisions
> > that may or may not have merges to/from other branches which also have
> > their own linear timeline.
> >
> > However, it then also says: "A branch might have other lines of
> > development..." which makes a branch sound more like an entire tree.
> >
> > Trying to reconcile these two statements.
>
> Technically, nodes in a tree have exactly one parent (zero if the root)
> and zero or more children, so that's the wrong model. The data structure
> is known as a "graph" which is like a tree but a node can have multiple
> parents.
>
> In an attempt to make Bazaar accessible to non-programmers, the
> documentation starts with the most common case - a linear timeline of
> revisions - and then attempts to build up to the true model (directed
> acyclic graph). Apologies for the resulting lack of clarity. If you can
> think of a better way of explaining this, please let me know.

Perhaps "A branch may pull-in / incorporate other lines of development...".  

I think a  way to avoid cognitive dissonance is to distinguish the unique
in-branch leftmost parent from tributaries.  An important property is that
though any node may have multiple parents -- all but the first being from
outside of the branch -- every node has at most one in-branch successor.

"Branches may split apart and be merged back together" is also a little
wishy-washy.  A single branch does not split.  It continues its linear
evolution.  A new branch may be cloned from an existing branch causing
it to duplicate all history up to that point.  This might be described
casually as "the branch split" but in reality it was duplicated so that
each branch might pursue a separate evolution.  Further, even when two
branches are reconciled to the point that their tips are identical they
have not "merged back together" to become one.  Each preserves its own
identity.  One branch may be abandoned because it is perceived as having
merge into another branch but that is a human interpretation and not a
property of the bzr data model.

/john