[Gnu-arch-users] Re: [BUG] Re: False "binary" files

[Stefan Monnier]

> Hmpf. But if we have an extensible set of ``diffing algorithms'' (which
> sounds very attractive indeed), do we have to care for `evolution'?
> Will we have to archive those algorithms as well? Or keep a central
> registry of blessed implementations? Or a combination of both?

These are only problems if you try to solve them at the wrong place.
Just move the responsibility to the archive maintainer.

> There seems to be some level of experience on those things among the
> distributed file system folks (Coda, Ficus).  They seem to use file-type
> specific conflict resolvers.  I have no idea of how they resolve (pun
> unintended ;) the above mentioned problem.

By delegating the decision to those who can take responsibility for it.


        Stefan