Re: [Gnu-arch-users] Google Summer of Code

[Thomas Lord]

Stephen J. Turnbull wrote:
>     Thomas> Arch is also still relevant compared to git because of
>     Thomas> integrity issues.
>
> By which you mean?
>
>   
Git picks a particular crypto hash and then pins all integrity on that.

That's too weak.

It'd take quite a few pages (forgive me for sparing myself from writing
at this particular instant) to explain in detail where I'm going with revc
but you need hashes, combined with signatures, combined with
auditing of reception, combined with open-endedness.   Archivists in
other domains are familiar with all this.  Git is just itchin' for 
trouble, as
it stands.   The only saving grace for git is that not enough infrastructure
has been built around it yet to make exploits worth more than ego 
gratification --
so far.


>     Thomas> You contrast git with arch by pointing out that git
>     Thomas> reconstructs (intrinsic) history from trees.  Um.  So does
>     Thomas> arch.
>
> Sure, but that's not what I meant.  The point is that in git because
> the name is a function of the semantics of the stored object, the
> "real history" and the "historical record" are one and the same.  If
> you find an object that is semantically identical, then it's part of
> the historical record, and has a natural place in it.  This is not
> true of arch; the history is an artificial cosntruct.
>
>   
No... damn.  Sorry to be abrupt but I think yr subtly confused.  I guess 
I need to do
a better job on explaining extrinsic v. intrinsic history.

-t