Proposal: Provide diffs between consensuses
nickm at freehaven.net
Sat Jun 28 04:26:29 UTC 2008
On Wed, Jun 25, 2008 at 08:30:09AM +0200, Peter Palfrader wrote:
> > 1. There needs to be a spec for generating and applying patches.
> > The spec for generating them needs to be deterministic (or have a
> > deterministic variant) so that authorities can all generate the
> > same patch.
> Why? Nobody needs to sign patches. Even directory caches can create
> the patches if we want. If all works out the end result will be a
> properly signed consensus.
Ah, you're right. I had assumed a signed-patch thing rather than a
patch-from-signed-to-signed thing. I take back the "deterministic"
part, but still mean the "specified" part. :)
> > 2. There should be a good Free Software library (3BSD-like license,
> > ideally) to generate and apply patches.
> Not that I have found one yet, but I didn't look too closely.
There are libraries for diff-related things in at least Python and
Perl, but I haven't yet turned up a C library with a license Tor can
use. I suppose we could extract the relevant parts of a 3BSD-licensed
"patch" implementation, but that way lies maintenance hell.
More information about the tor-dev