[Starlink] fcc NOI response due Dec 1

[Sebastian Moeller]

Hi Alexandre,

> On Dec 7, 2023, at 12:49, Alexandre Petrescu via Starlink <starlink at lists.bufferbloat.net> wrote:
> 
> Thank you for having prepared this response.
> 
> It is a US-centric context, but it might apply everywhere else where fiber and satcom access are considered in competition.  Besides, the latency reduction priming over bandwidth increase, might be discussed in a 6G context as well, be that with NTN or without.
> 
> Now that the we are past the deadline, I would like to mention two other aspects:
> 
> - all access kinds in recent years have witnessed combined improvement of bandwidhts and latencies.  Within a same access kind (e.g. within WiFi, within Ethernet, within cellular) each increase of bandwidth was accompanied by a decrease of latency.  As such, it might look surprising to argue in favor of latency decrease at the expense of a constant bandwidth.  It might not happen, because traditionnaly they are combined.

	[SM] Except where they are not... looking at access networks sure coming from acoustic couplers in double/triple digit Baud range over analog modems in the 56Kbps range and ISDN and DSL access latency generally decreased, but DSL already allows sub-millisecond access networks to be built (typically that was not done as interleaving resulted in considerably more robust networks). And with the roll ouf of G.INP retransmissions noticeable interleaving essentially was removed, but this resulted in downstream access speeds ranging from (to cite local numbers) 16, over 100 and 250 Mbps all operating with the same access latency (based on DSL's 4KHz "clock")... Similarly DOCSIS networks have been ramping up from single digit Mbps numbers to well beyond 1000 Mbps all with similar ~2-4ms request-grant latencies (I might be wrong here and this might only apply for DOCSIS >= 3.0, I am sure we have real experts on list that will correct anf misconception). And IMHO the same holds for GPON/XGSPON where the 8KHz request-grant clocking results in similar ~2ms access delays (in upload direction).
	So if looking from far above, sure in the course of the last 30-40 years both throughput increased and latency decreased, but once we look closer there does not really seem to be an automatic inverse proportionality between these two measures... To give an example loading a semi-trailer in NYC to the bring with harddisks and drving these to say LA will have atrocious latency but the throughput will still be decent...


> - a strong argument could be made in favor of satcom over fiber in remote areas: satcom avoids the tangled fibers and satcom might pollute less than fiber; but at two conditions: satcom should have a sat exit strategy (more than just burning upon re-entry, maybe more recover and reuse, less visual pollution with maybe more paintings) and (2) satcom should aim at a same kind of... latency (yes, that!) that fiber aims at.  The 10ms that starlink aims at is way too high compared to what fiber access latency aims at.  IT is possible to aim at lower.

	[SM] End-user fiber access is typically implemented as shared passive optical network (PON) for cost/profitability reasons and these are not orders of magnitude better than 10ms, so at 10ms satellite based-internet access will well be competitive. However personally I think we should strive for all fiber everywhere and use things like satellite internet as bridge technologies and/or emergency systems or mobile access in remote areas.  Not because going via space is not inherently impressive (and might I say cool?) but more because I assume the internet is here to stay and we are talkning about infrastructure that will be in use hopefully for decades.

Regards
	Sebastian



> 
> Alex
> 
> Le 27/11/2023 à 16:53, Dave Taht via Starlink a écrit :
>> We started work on a response to the FCC NOI requesting feedback as to
>> future broadband bandwidth requirements for the USA early this
>> morning.
>> 
>> I am unfamiliar with the processes by which Starlink was disqualified
>> from the RDOF?, and a little out of date as to current performance. It
>> is very clear they are aiming for 100/20 speedtest performance and
>> frequently achieving it.
>> 
>> A drafty draft is here, and some of the language is being toned down
>> by popular request. (the pre-readers were lucky! I cut the cuss-words
>> out) There is only one joke in the whole thing. I'm slipping!.
>> 
>> https://docs.google.com/document/d/19ADByjakzQXCj9Re_pUvrb5Qe5OK-QmhlYRLMBY4vH4/edit?usp=sharing
>> 
>> I have some starlink info contained in appendix B so far, but I would
>> prefer not to cite my own long term plot as I did, and  also cite
>> others that have a good latency measurement, I like the 15s irtt plots
>> I have seen gone by. If you have research about starlink you would
>> like me to cite in this context, please comment on the link above!
>> 
>> The NOI is the first link, and it helpe me, actually, to start with
>> the FCC commissioners' comments at the end, rather than read through
>> the whole thing. Not that I would not welcome more folk submitting
>> themselves to that...
>> 
>> 
> _______________________________________________
> Starlink mailing list
> Starlink at lists.bufferbloat.net
> https://lists.bufferbloat.net/listinfo/starlink