incubator-wave-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Bruno Gonzalez (aka stenyak)" <>
Subject Re: IRC discussion on P2P waving
Date Fri, 21 Jun 2013 14:48:29 GMT
On Fri, Jun 21, 2013 at 3:06 PM, John Blossom <> wrote:

> On the P2P latency, I don't think that it would be unacceptable to draw a
> line and say that P2P provides limited, non-guaranteed realtime OT or that
> it's not realtime OT and more of a syncing mode than a conversation mode.
Rather than unacceptable... it might simply be unavoidable! :-)

> That would probably be sufficient for what needs to be done, especially
> since in some instances P2P-enabled Wave sessions may be using MESH
> networks for transport - a key factor in how a lot of experimental
> communications services are being deployed in developing nations (not just
> the Project Loon concept). In the MESH model, you're likely to have one
> node within range of another temporarily, which may sync with it, and then
> pass along data to another node when it comes in range of it. That's the
> most probable scenario for P2P in many instances, I would think. The other
> potential scenario: two people in a remote location, for the sake of
> argument two movie script-writers who have holed themselves up in a remote
> location to collaborate on a common script. They're on two devices that are
> very proximate to one another, so perhaps the latency issues will not be so
> severe.
To make it clear:
The use of very laggy DHT-like protocols only affects in use cases where
the peer search space is huge. Such as the world wide Internet.

In other situations (your average LAN, a bluetooth and IR connections, etc)
DHT would not be necessary, because other techniques exist that allow to
discover peers more effectively (dns-sd, bonjour, plain broadcasts to the
whole subnet...). These techniques would not have the huge delay associated
with DHT, and therefore would allow real time waving. This is already used
by other P2P systems such as BitTorrentSync.

Taking your use cases:
 - Real time waving would be definitely be possible between the 2 script
writers, for example (we assume they can use a local link to each other, so
they don't need to go through the whole internet to find and communicate
with each other).
 - I don't know if Project Loon would allow connections between close
peers. If it works like your typical ISP, they need DHT even if the peers
are 2 houses away from each other, therefore real time is not doable. If it
works like a wifi access point (which happens to be in the sky, rather than
on your living room), then it would be similar to a LAN, DHT could be
avoided, and real time waving would be possible.
 - I have no knowledge about the "MESH" network so I can't comment on that.

But I agree with Pratik and Joseph: let's start simple, try to get the git
use cases right, and we have the simple stuff working, then we can worry
about doing the advanced stuff that nobody has done before.
For this, we need a P2P OT (even if we end up all using yahoowave and
googlewave servers, instead of pure p2p connections).

     Bruno González

Jabber: stenyak AT

  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message