Re: OOB Reply Delivery Protocol v3

Zlatin Balevsky wrote:
> >No vendor is
> >allowed to utilize any of them without a prior announcement and documentation.
 
> The BitTorrent protocol reserved 64 bits for protocol extensions with
> exactly the same restriction.  The result was massive abuse, chaos and
> lack of consistency.

Reminds me of Gnutella. However who do you expect to introduce this chaos? How
many people are working on Gnutella software at this point?

> The good thing about ggep keys is that they are in the alphanumeric namespace
> and there is no "next" key a vendor can decide to loot.  It is easy to avoid
> collisions even without central naming authority by prefixing or a naming
> scheme.

That's exactly how it works. That's why I introduced GTKG.TLS and GTKG.IPV6
instead of TLS and IPV6. Any GGEP IDs shorter than 5 characters are reserved.
Vendors are free to use whatever GGEP IDs they want as long as prefix them
with their vendor ID. These have been the rules since day one of GGEP.

I don't have a strong preference for GGEP F though. If the current specs are
final, my work is finished. Gtk-Gnutella supports OOB v3 completely. We might
want to agree on some flag day at which point OOB v2 will be no longer
supported. As Gtk-Gnutella expires in one year, I'd consider a similar time
frame appropriate.

-- 
Christian