[Starlink] IXPs in space

[David Lang]

nobody else has had enough satellites in space to do real in space routing, so 
take anything you read about it that's not specifically referring to starlink 
as guesses/suggestions.

David Lang

On Mon, 17 Apr 2023, David Fernández via Starlink wrote:

> Date: Mon, 17 Apr 2023 16:22:33 +0200
> From: David Fernández via Starlink <starlink at lists.bufferbloat.net>
> Reply-To: David Fernández <davidfdzp at gmail.com>
> To: starlink <starlink at lists.bufferbloat.net>
> Subject: Re: [Starlink] IXPs in space
> 
> Apologies in advance for any *profound* misunderstanding that could
> trigger any nerves, although I am grateful for all the info being
> shared. I will read this carefully:
> https://web.mit.edu/Saltzer/www/publications/endtoend/endtoend.pdf
>
> In the abstract, I read, at the end: "Low level mechanisms to support
> these functions are justified only as performance enhancements." So,
> every rule has its exceptions, like the PEPs for TCP performance
> acceleration.
>
> "Adding those anycast addresses to the satellites would be transparent
> to all users (assuming the satellites are operating at the IP layer,
> the old bent-pipe approach did not, but once you have routing in space
> via the laser links...)" Satellites being routers just because they
> have ISL does not follow. Is there any satellite nowadays doing
> routing in space? They do switching only (and not Ethernet switching).
> Because of this I proposed the idea of the so called L2 snooping. It
> is a hack, somehow similar to the PEPs to accelerate TCP, but in this
> case to accelerate DNS (not encrypted), without making satellites
> routers, keeping them transparent as they are now.
>
> If it has not been done before by anybody is for a good reason, maybe.
> Things like this did not have so much success, it seems:
> https://en.wikipedia.org/wiki/Internet_Routing_in_Space
>
> "DNS is an easy thing to start with compared to most others." Even in
> this case you are going to find a lot of issues to do it with
> transparent satellites, even with GEO satellites, where the gain could
> be bigger and things simpler than with fast moving LEO satellites.
>
> "bent pipe for normal operations doesn't mean that you can't watch for
> an anycast address to serve locally." This is what I was suggesting,
> you sniff packets in the satellite uplink and answer anycast DNS
> queries directly from satellite, if you see any, cutting RTT by a
> half.
>
> Regards,
>
> David
> _______________________________________________
> Starlink mailing list
> Starlink at lists.bufferbloat.net
> https://lists.bufferbloat.net/listinfo/starlink