More merge base discussion [was: monotone's LCA+DOM algo for selecting a merge base]

[Denys Duchier]

I have been thinking about this issue just a bit more, and I am beginning to
realize the value of starting with scalar merge (well... duh!).  It certainly
eschews the complexities of hunk-based conflicts.  Things are a lot crisper when
there is just one scalar value per revision.

I'll probably be offline for the rest of the week, but let me at least offer a
couple of definitions for the scalar case:

ISOPLETH COMPONENT:

an isopleth component of a revision graph is a maximal connected set of
revisions all having the same scalar value.

ISOPLETH QUOTIENT GRAPH:

the graph obtained from a revision graph by collapsing all revisions in the same
isopleth component.

I'll try to flesh out my previous argument in this context.

Cheers,

--Denys