Re: [Starlink] Starlink tidbits from NANOG

[Darrell Budic <budic@onholyground.com>]

> On Nov 4, 2021, at 11:46 AM, Dave Taht <dave.taht@gmail.com> wrote:
> 
>> - v6 is deliberately not fully functional, but they know some of use are using it and it will eventually be fully activated. May be waiting on the regional connectivity, so will be intersting to see if changes for some areas and not others as they roll it out.
> 
> Deliberate... well, if you ship 10 year old openwrt software to users
> (we'd made a big push for ipv6 there before ipv6 launch day in 2013),
> and
> don't keep up...  I guess you could call that deliberate. I'm pretty
> happy with openwrt 20.2.1. IMHO: ipv6 really requires a modern kernel
> and tools, not less than 4 years old, to deploy well. Maybe there's a
> worthwhile SDR stack, I don't know...
> 

Deliberately not enable radvd in this case, so hopefully easily solvable once they have a v6 backbone they are happy with.

>> - They hate Google's outsourced NOC as much as the rest of us
> 
> Do say more. :) I'm told google at least have a very nice set of tools
> for looking at the characteristics of interchange traffic.

If you’ve ever had to deal with them, it’s a lot of long delays on tickets, no explanation for problems and fix acknowledgements only if you’re lucky, and lots of “please do the needful and revert…” type non-answers. And they never give out data on your traffic once it’s left your box/boundary, even though you’re paying them for the traffic. One of the worst I’ve ever delt with, but fortunately offset by all the smart folk you never get to talk to that ensure most things work really well most of the time.


>> - New ground stations with more capacity are coming (and will be upgrades).
> 
> Would so love universities to get in on some of those. I remember the IMP.

They are emphasizing public peering, so I’m hopefully they’ll keep moving. Many of the university’s already peer at SIX and a couple are at ChIX, so their connectivity will open up a bit. Speaking of, I’ll give a 10G ChIX connection to any public educational institution that wants one and can get to me, so send ‘em my way if they’re in the midwest and interested!

> 
>> - new birds also have 2-3x more ku bandwidth than first gen
> 
> up and down?
> 
> My take on the up problem was that it was regulatory. ?
> 

It’s three beams “up/down” from the Dishys (one at a time in a TDM fashion, which we knew), and one locked on whatever ground station it’s tracking at the moment. Sounded like more capacity on all of them, unclear if it’s better antennas, more power, better tracking, or all of the above. It did sound like the ground stations will get bigger antennas, so that’s probably part of the downlink improvements.

> 
> Just someone telling me under pain of death, "dave, you can't talk for
> X months, but we're going to do cake/fq-codel/pie/something" would be
> comforting. There's a whole internet elsewhere left to fix, starlink
> getting it right and a little publicity around it would do wonders...
> and certainly wifi is highest on my list. As it is, I got annoyed
> enough last week to try and get the autorate sensing code to work well
> on starlink. There's a prototype now that seems to be working well on
> lte, see here: https://forum.openwrt.org/t/cakes-autorate-ingress-testing-needed/108848/186
> 
> Testers wanted.

I’ll get my cake install upgraded and test it out. To be honest, I haven’t even been running it for a while, so they’ve gotten better since you started talking to them even if it isn’t great yet.

> Fixing fixed wireless has been a pain point far, far, far greater
> than the disappointment I felt at starlink so totally missing the
> bufferbloat problem initially. It will take a decade to sort out 5g,
> 4-6 months more for starlink oh! yes! yes! yes!
> 
> More news on that as it happens.
> 
>> - it’s encrypted up and down. I didn’t know that yet, but I may have just missed it.
> 
> I did. But it's really hard to trust that black box and the world has
> otherwise shifted to e2e encryption.

My contact acknowledged that, but it’s nice to know it’s probably better encrypted than the LTE modems we all use everyday already. I imagine we’ll have to continue to depend on the application layer for that sort of e2e encryption, with the advantage that it won’t matter what our transport tech is then.