Hacker News Reader

Comment by gnode | Hacker News Reader

I wonder if the reason here is that the topology of the network will change rapidly. E.g. as equatorial and polar satellites cross each other, they will be in direct communication with each other for only brief periods of time. There are ways to cope with routing changes in IP, but I don't know if it'd be seamless enough.

View on HN

It's easy to read too much into a tweet, but it sounds like the header size overhead and number of options are the primary reasons for not using IPv6.

Note also that IPv4 and IPv6 are nearly independent of the routing protocol.

Also, my first job was writing discrete event models of wireless communications equipment, including a couple satellite models. You'd be surprised how simple satellites themselves are. A lot of work goes into moving the complexity from the fragile, expensive, and out-of-reach satellites and into the ground stations. I wouldn't be surprised if packets weren't at least partially source-routed, which might very well involve broadcasting GPS-like tables of satellite orbit parameters.

I expect that internet traffic will still travel over the network using IP, but encapsulated in the SpaceX protocol. I expect translating the headers to and from IP would not make the protocol "simpler than IPv6", and would not be justifiable complexity to save a few bytes.

I think this "simple" protocol is just to simplify routing, and allow non-IP traffic over the network (in-band control messages; phone calls?).

Looking back at the context, I think you're right. It looks like Elon was just trying to explain to someone without much networking background that, like most link layers, they'd be using a link layer more simple than IPv6. He probably wasn't making a statement about lossy header compression that discards uncommonly used features.

Comments up-thread speculate Musk may be looking at something like ATM, which uses small, fixed packet sizes for better flow control, along with expectation packets may arrive out of order due to traversing multiple paths. ATM has implementations for reordering packets, and the fixed packet size ensures reordering can be accomplished within a fixed latency.

Reply on news.ycombinator.com