Re: [Monotone-devel] partial pull #2 - gaps instead of a single horizon

[Markus Schiltknecht]

Hi,

Justin Patrin wrote:
> Except that, as has been said many times, there is no real third-party
> time.

Well, netsync would probably fail to optimize away unneeded revisions if 
the clocks of the two peers are more that a day apart. In the worst 
case, that should result in a larger merkle trie than really necessary. 
Not that much of a problem, IMO.

> While the certs for revs *should* have the same time this isn't
> necessarily true. Of course this would probably mean we just transmit
> a few extra certs or something.

Yup.

Perhaps it's better to limit by depth, i.e. pulling the last 10 revisions.

For sync, it's probably not even necessary to give a limit. Instead both 
peers should exchange their gap starts and endings prior to building the 
merkle trie - which they can then build depending on what gaps they want 
to fill.

Honestly, I didn't think too hard about the netsync part of the story. 
I'm trying to get a local database and workspace flying, despite having 
only parts of the repository. As soon as that works, we can talk about 
netsync and optimizing it for partial repositories.

Regards

Markus