[darcs-users] darcs patch: resolve issue27: add junk to patch identifiers.

Lele Gaifax lele at nautilus.homeip.net
Thu Sep 18 06:52:49 UTC 2008


On Wed, 17 Sep 2008 14:44:50 -0400
David Roundy <droundy at darcs.net> wrote:

> On Wed, Sep 17, 2008 at 07:41:04PM +0100, Eric Kow wrote:
> > To justify this, I could invoke a belief in Murphy's Law, i.e. that
> > if anything could get Tailor users up in arms, it will get Tailor
> > users up in arms.  (For the interested, the "anything" in question
> > is that tailoring the same repository twice will now produce two
> > subtly different repositories)...
> >
> 
> I think deterministic junk is considerably harder to produce, and
> requires proof that it is generated correctly, while non-deterministic
> is easy to produce.  We can always change it to be deterministic
> later.

I'm still convinced this is wrong, or better, nobody convinced me
about the usefullness of adding random junk to the identifier
computation. On the tailor side, I tried to make it evident in the
documentation and used a (relatively) safe default for the involved
option, "patch-name-format": I'm very sorry if that's not enough.

OTOH, David already explained that any alternative (such as detecting
the problem in advance) is very hard to impossibile to do: I can't
predict whether the new behaviour of darcs will be any safer, but I'm
biased to think that there should be an option to control the random
injection, so that a person who knows what he's doing could still mix
patches coming from two dictinct tailorizations of the same source (I
never had the need myself, however).

ciao, lele.
-- 
nickname: Lele Gaifax    | Quando vivrò di quello che ho pensato ieri
real: Emanuele Gaifas    | comincerò ad aver paura di chi mi copia.
lele at nautilus.homeip.net |                 -- Fortunato Depero, 1929.


More information about the darcs-users mailing list