On Mon, Sep 30, 2019 at 11:15:46PM -0400, Keith Moore wrote: > On 9/30/19 5:36 PM, Bob Hinden wrote: > > 2) The second reason is that I think the reason for few IANA > > allocation requests in this registry is that it is likely that > > packets containing any new assignments will be blocked in firewalls > > and middle boxes. It’s hard to get a new protocol deployed. I am > > doubtful this will change anytime soon. I suspect we will never > > run out, unless the Internet changes significantly. > > I'm always amused, and a bit concerned, by the assumption that the Internet > will not change significantly. From my perspective, the Internet seems to > constantly change. Not long ago it was hard to imagine any new transport > protocol getting much traction, for example. And the tussles between the > application writers and enterprise network operators haven't gone away. The Internet could change so much as to disappear, or so little as to be recognizable in 30 years. Who knows. But the forseeable future as to new transport protocols atop IP is pretty easy to forecast: as long as default middle box behavior is to drop protocols they don't speak, new transport protocols won't get deployed. And these middleboxs... many of them are cheap, update-less consumer devices that will be with us for many years yet. On the plus side, there may be no need for NAT for IPv6, on the minus side, firewalls are still a selling point, and NAT is part of it. Perhaps we could just have a common header for new transports that bears port numbers so that middle boxes can NAT them without knowing anything about them... Still, we'd be at least half a decade from being able to deploy new transports. Maybe a killer app could materialize that makes consumers upgrade their middleboxes? Nico --
