From: Michael S. Tsirkin <mst@redhat.com>
Date: Tue, 17 Sep 2013 20:41:59 GMT
Subject: Re: [PATCH] diff: add a config option to control orderfile
Message-ID: <20130917204159.GA22116@redhat.com>
In-Reply-To: <20130917203807.GA22059@redhat.com>

On Tue, Sep 17, 2013 at 11:38:07PM +0300, Michael S. Tsirkin wrote:
> On Tue, Sep 17, 2013 at 04:18:28PM -0400, Jeff King wrote:
> > On Tue, Sep 17, 2013 at 11:16:04PM +0300, Michael S. Tsirkin wrote:
> > 
> > > > Thinking about it some more, it's a best effort thing anyway,
> > > > correct?
> > > > 
> > > > So how about, instead of doing a hash over the whole input,
> > > > we hash each chunk and XOR them together?
> > > > 
> > > > This way it will be stable against chunk reordering, and
> > > > no need to keep patch in memory.
> > > > 
> > > > Hmm?
> > > 
> > > ENOCOFFEE
> > > 
> > > That was a silly suggestion, two identical chunks aren't that unlikely :)
> > 
> > In a single patch, they should not be, as we should be taking into
> > account the filenames, no?
> 
> Right.
> 
> > You could also do it hierarchically. Hash each chunk, store only the
> > hashes, then sort them and hash the result. That still has O(chunks)
> > storage, but it is only one hash per chunk, not the whole data.
> 
> Could be optional too :)
> Or maybe just sum byte by byte instead.

One's complement probably ...

> > A problem with both schemes, though, is that they are not
> > backwards-compatible with existing git-patch-id implementations.
> 
> Could you clarify?
> We never send patch IDs on the wire - how isn't this compatible?
> 
> > Whereas
> > sorting the data itself is (kind of, at least with respect to people who
> > are not using orderfile).
> > 
> > -Peff