Workflows, rebase, patch theory

[Andrew Bennetts]

Stephen J. Turnbull wrote:
> Andrew Bennetts writes:
> 
>  > "Set aside" or "ignored" are perhaps more accurate terms  [...].
> 
> Yes, they are.  Isn't that good reason to prefer them?

Well, they didn't occur to me when I wrote the original mail in this thread that
first used “thrown away”.  Because:

>  > From the perspective of the rebased branch, that history is thrown
>  > away.

That was the perspective I was working from.  The context I was writing about
was the branch.  What happens elsewhere is irrelevant to discussing what I was
talking about.

> Sure, with "perspective" being the operative word.  Ie, it's purely a
> presentation issue: the rebased branch presents the changes cleanly
> relative to the merge base.
> 
> This is a very useful presentation; it is typically what a gatekeeper
> wants to see.  The original history is still available to the
> submitter if she needs it, and can easily be made available to others.

I've never disputed the usefulness of this presentation.

However, rewriting recorded data merely to get a prettier presentation strikes
me as suboptimal.  I don't believe creating this presentation is mutually
exclusive with using the original history directly.  Looms are an example of an
alternative approach to this problem that doesn't require synthesising a new,
conflicting history.

-Andrew.