[Starlink] Starlink tidbits from NANOG

Michael Richardson mcr at sandelman.ca
Thu Nov 4 15:10:59 EDT 2021


<#secure method=pgpmime mode=sign>

    >> Even if they used IPv6 for the "last mile" and ran NAT64, that would still be
    >> a major improvement over trying to run dual stack.   That's what the smarter
    >> mobile operators are already doing. (jio, some US carriers...)

    > have these operators have solved the mobility problem ipv6 has always
    > had?

If you are speaking about the issue that Enterprises have that they'd like to
continue to run their internal network when changing ISPs (using NAT44 at the
edge), then the answer is that mostly, it's a problem caused by IPv4-think.
(The fact that public /48s are not yet free is also an issue)

I don't expect that Starlink will have this problem, and in the home, the
ULAs generated by 7204 compliant CPE devices solves most issues.

(We still have a source address selection problem: but that problem won't go
away until a few dozen core Google/Microsoft/Apple developers are forced to
live in an IPv6-only home network.)

And JIO (India) and other LTE operators' deal mostly with handsets, which
don't have permanently up internal networks (alas).

    > if not then I tend towards a custom l2 that supports anything on top
    > of it, which might
    > even include cool things like icn.

More L2 tricks will not save the day. It just makes the network invisible and
thus undebuggable.   It's time to acknowledge that ethernet is not a fat Yellow COAX.

--
]               Never tell me the odds!                 | ipv6 mesh networks [
]   Michael Richardson, Sandelman Software Works        |    IoT architect   [
]     mcr at sandelman.ca  http://www.sandelman.ca/        |   ruby on rails    [



More information about the Starlink mailing list