[NNagain] RFC: Public Communications on Tech Infrastructure

Wed Oct 25 09:26:57 EDT 2023

Branching from Dave's thread because I don't want to get into the politics,
but I would like to very strongly endorse Dave's remarks about how the
incredible accomplishments of network engineers are totally unacknowledged
and misunderstood (e.g., note the public policy emphasis on line speed over
all else.) As such, I'd like to solicit the members of this list to suggest
some of the greatest accomplishments in network engineering that you've
never seen properly acknowledged or appreciated. I'd like to promote and
discuss them in speeches and papers to help get more sunlight on them.

*0. Let's get network engineering some applause, please!*
Both recent and historical accomplishments are welcome. I just want to help
get more people thinking about what a difference network engineering has
made to everyone's lives! All technologies, personalities and
accomplishments welcome!

Beyond this specific thing, in terms of public discourse, I'd love to get
more opinions about how to communicate to the public about the tech
underpinnings of the world we live in now, and I'd love comments on how to
discuss and promote any of these topics:

*1. Infrastructure advances*
It would generally do a lot of good if the public were to think of "tech"
less as purely the consumer-facing side and more in terms of fundamental
architecture and infrastructure. For example, there's really no point
talking about "AI" in the public-facing aspect of end-user LLM experiences
without first looking at how the cost of compute and transit has gone
through the floor compared to 15 years ago or so. I can't even disentangle
all the drivers, but they must include at least:

   - New uses for GPUs driving advances and slashing prices in GPU tech
   - Vast advances in back-end cloud (to pick one company,
   Sawzall/Lingo/GFS/Colossus plus associated datacenters is almost invisible
   to the public, and I have no idea what's powering Chinese AI back-ends)
   - Nuts-and-bolts development in ML/data science that are eroding the
   fuzzy boundary between ML done as a planned, discrete query by an expert
   over a small, curated dataset and ML as a quasi-autonomous system not
   requiring expert queries, given authority over physical devices, doing its
   own ingestion, etc -- "a sufficiently large difference in quantity is
   itself a difference in quality"

This stuff is particularly worth asking about because we are now at least
30 years into what I think of as "pervasive networked personal computing,"
now in wireless and appified form, and I think the public experiences this
as just advances that "happen by themselves" in the ordinary course without
seeing the jags in the step functions underwriting the apparent smooth
curve of progress.

*2. Security in real-world systems*
Getting hacked used to mean losing data, having devices bricked, maybe
getting co-opted into a botnet, etc. Now it's a lot scarier, because we are
increasingly surrounded by always-on, always-connected devices whose
security infrastructure is a black box and which may be trusted with
controlling physical equipment. It's bad enough if your household
appliances are phoning home (where?) with your credit card number. It's a
whole new level of scary if there are possible APTs in the power grid and
whoever manufactured the IOT modem in a transformer is about 8 degrees of
separation from the grid operators. Even if there's no malice intended,
modern grid balancing is a new level of challenging because you may have
multiple sources of generation with immense moment-by-moment fluctuations
in inbound generation, etc., and that's just one category, leaving
groceries, ports, financial markets, building security, whatever replaces
positive train control (PTC) down the road, vehicular autonomy, industrial
operations, etc. to one side...

Panic reactions are one thing, but it would be more productive for the
public to think about what their expectations are for how to react to these
new capabilities and challenges and then demand that the policy sector
cashes this out into new standards by consulting with technologists. I
would therefore love advice on what you think the public needs to know.
Maybe some kind of public forum that could get press or a white paper that
could get written up in an op-ed?

On that note, in addition to (or instead of) commenting on this posting,
please consider commenting on the US Cyber Trust Mark proceeding now open
at the FCC (comments close November 10th, commenting link here:
https://www.fcc.gov/ecfs/search/docket-detail/23-239). If you'd like to
talk about this off-list, please drop me a line at NS at fcc.gov. I'll let you
know in advance if anything you want to say requires you to file an "ex
parte" statement so that you don't have to worry about going on the record
unintentionally. This is a fantastic opportunity for the network
engineering and computer security communities to air their concerns in a
federal forum in a way that may bind the federal government going forward.

*3. The future isn't evenly distributed*
Talking to a friend who does industrial devops reminded me of this
fantastic postmortem on healthcare.gov's rollout:
https://lobste.rs/s/igt4ez/10_year_anniversary_healthcare_gov. Obviously I
don't need to tell the career professionals this, but tech advances don't
necessarily propagate, and if they do, it may be at radically different
rates between different countries, companies, sectors... (If I needed a
reminder of this, I recently had to upload DICOM files to a hospital using
a terrible Java applet that was obviously written so long ago that it only
wanted to upload from CDs, i.e., at a time when you wouldn't have spent
hard disk space on DICOMs. I eventually managed to "persuade" it that a
flash drive was a CD.)

This ties into points 1 and especially 2, because if we want the full
social benefits of all the advances modern engineering has accomplished, we
need to get people in "nontraditional" sectors thinking about the benefits
of the communications and controls capabilities that are now on the table.
Everyone should be asking why we aren't doing ML to reduce the cost and
energy consumption for making breakfast cereal, totally pedestrian stuff
like that; if the answer is juice isn't worth the squeeze, that fine, but
that's going to run on a delay because, as the healthcare.gov example
shows, high-value new practices may be invisible to a sector that would
definitely benefit from them.

Sorry for the very lengthy post, and as they say on the artist formerly
known as Twitter, "my DMs are open." And thanks for everything you all do!

All the best--
Nathan

On Tue, Oct 24, 2023 at 3:22 PM Dave Taht via Nnagain <
nnagain at lists.bufferbloat.net> wrote:

> On Tue, Oct 24, 2023 at 11:21 AM the keyboard of geoff goodfellow via
> Nnagain <nnagain at lists.bufferbloat.net> wrote:
> >
> > ➔➔https://twitter.com/BrendanCarrFCC/status/1716558844384379163
>
> Leaving aside the rhetoric, I believe the majority of these claims on
> this part of his post:
>
> https://twitter.com/BrendanCarrFCC/status/1716884139226329512
>
> to be true. Any one question this?
>
> I do wish that he showed upload speeds, and latency under load, and,
> acknowledged some mistakes, at least, and did not claim perfect
> success. Also individual states had stepped up to institute their own
> rules, and I would love to see a comparison of those stats vs those
> that didn´t.
>
> The COVID thing I am most fiercely proud of, as an engineer, is we
> took an internet only capable of postage stamp 5 frame per sec[1]
> videoconferencing to something that the world, as a whole, relied on
> to keep civilization running only 7 years later, in the face of
> terrible odds, lights out environments, scarce equipment supplies, and
> illness. ISPs big and small helped too - Their people climbed towers,
> produced better code, rerouted networks, and stayed up late fighting
> off DDOSes. People at home shared their wifi and knowledge of how to
> make fiddly things on the net work well, over the internet  -
>
> Nobody handed out medals for keeping the internet running, I do not
> remember a single statement of praise for what we did over that
> terrible time. No one ever looks up after a productive day after a
> zillion productive clicks and says (for one example) "Thank you Paul
> Vixie and Mokapetris for inventing DNS and Evan Hunt(bind)  and Simon
> Kelly(dnsmasq) for shipping dns servers for free that only get it
> wrong once in a while, and then recover so fast you don´t notice" -
> there are just endless complaints from those for whom it is not
> working *right now* the way they expect.
>
> There are no nobel prizes for networking.  But the scientists,
> engineers, sysadmins and SREs kept improving things, and are keeping
> civilization running. It is kind of a cause for me - I get very irked
> at both sides whining when if only they could walk a mile in a
> neteng´s shoes. I get respect from my neighbors at least, sometimes
> asked to fix a laptop or set up a router... and I still share my wifi.
>
> If there was just some way to separate out the ire about other aspects
> of how the internet is going south (which I certainly share), and
> somehow put respect for those in the trenches that work on keeping the
> Net running, back in the public conversation, I would really love to
> hear it.
>
> [1] Really great talk on networking by Van Jacobson in 2012, both
> useful for its content, and the kind of quality we could only achieve
> then: https://archive.org/details/video1_20191129
>
> > --
> > Geoff.Goodfellow at iconia.com
> > living as The Truth is True
> >
> > _______________________________________________
> > Nnagain mailing list
> > Nnagain at lists.bufferbloat.net
> > https://lists.bufferbloat.net/listinfo/nnagain
>
>
>
> --
> Oct 30:
> https://netdevconf.info/0x17/news/the-maestro-and-the-music-bof.html
> Dave Täht CSO, LibreQos
> _______________________________________________
> Nnagain mailing list
> Nnagain at lists.bufferbloat.net
> https://lists.bufferbloat.net/listinfo/nnagain
>

-- 
Nathan Simington
cell: 305-793-6899
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.bufferbloat.net/pipermail/nnagain/attachments/20231025/4c60b214/attachment.html>