Pushing after merge considered harmful

[Stephen J. Turnbull]

Martin Pool writes:
 > 2010/1/26 Stephen J. Turnbull <stephen at xemacs.org>:

 > > That reads like an indictment of the Bazaar UI to me.  "There's no
 > > complete description of the semantics of Bazaar commands, because it
 > > would be huge and seem full of special cases (and always be
 > > incomplete, too)."  Since a_r_o affects only a subset of the Bazaar
 > > UI, you are a fortiori making that broader claim, too.
 > 
 > This option works at the level of the dag model: there is no
 > per-command implementation.  We explained what effect it has there,
 > and how the commands map to the dag.

That resolves to a claim that special cases are neither many nor
incomplete.  I don't have a problem with that, but it contradicts
Matthew.

 > (I think the latter at least could be far far better in the manual
 > though.).  Explaining the cross product of every dag-level thing
 > and every ui level thing seems O(n^2)

Which is actually O(1) if you put a strict bound on n. ;-)  But even
taking into account your implied intent to proliferate the DAG API and
the CLI commands without bound :-), it's O(n + n).  Eli would surely
be happy if the docs explained the DAG level functionality well, and
gave a precise, DAG-level description of what commands do.

 > and just the kind of thing you would criticise us for doing.

I reject that claim.  I would criticize you for proliferating
features, not for writing good documentation to match the features you
provide.

 > > <snort>  I can't speak for Eli, but I certainly have no trouble
 > > understanding that.  However, the Moebius strip being one of my
 > > favorite toys, "append revisions only" sure is a name that does
 > > nothing for my intuition about the semantics of the option.  Maybe I'm
 > > just weird, though.
 > 
 > Under what name would you expect to find it?

"preserve_mainline" is most accurate, I guess, but if I were looking
to solve the problem reported by the Emacs developers, I'd probably
start by looking for "preserve_parent_order".