[darcs-users] Re: What's better, Darcs or Mercurial?

[Tommy Pettersson]

On Tue, Feb 28, 2006 at 09:27:17AM +1100, Tim Docker wrote:
> Bryan O'Sullivan wrote:
> 
>  > The big drawback of darcs is that it has a "theory of patches" :-)
>  > In many practical cases, it simply falls on its face and takes
>  > seemingly exponential time and space to try to figure out the
>  > interrelationships between patches. My impression is that there
>  > is no sign of these problems being resolved any time soon; they
>  > appear to be a fundamental side effect of the current design.

I don't know where patch theory stands right now, but I share
the belief (or at least suspicion) that it may always be
possible to construct a merge situation fundamentally
exponential to solve.

> Interestingly no-one replied to this assertion. Do darcs developers feel
> that the problem of "poison patches" etc is one that will be resolved
> in the near future?

Another way to look at it is that darcs _allows_ you to create
"unsolvable" problems. After learning how to think like darcs it
is not very hard to avoid the exponential pitfalls. Currently
this is (unfortunately) a bit like learning how to ride a
bicycle.

More regrettable this leads to darcs not being as usable as
intended in some common usage patterns because you tend to avoid
them all together although they could work fine as long as you
don't happen to get the bad kind of conflicts. I think the new
conflictors are meant to remedy much of this because bad
conflicts in these usage patterns are some how conflicts with
other conflicts that could be sorted out, sort of, I think.


-- 
Tommy Pettersson <ptp at lysator.liu.se>