IP over Lasers
mikekohn.netThey built a program they called (relay) that sends packets between a tun network device & a uart serial device
https://github.com/mikeakohn/small_projects/blob/main/ip_ove...
I .... wonder if they considered just using PPP/SLIP?
A while back I was doing some IP-over-UART shenanigans and it took me a surprising amount of research to discover the existence of SLIP. I think it's one of those things that at some point "everyone knew about" and ended up being poorly documented as a result.
SLIP is old and mostly forgotten for sure, but PPP was almost universal in dialup and unfortunately common in DSL. PPPoE was the bane of my existence as someone who had to support third-party routers on DSL. PPTP was also very common for small VPNs even well past when it should have been retired due to insecurity because it worked out of the box on Windows and Mac without a third party client.
Mid-90s you definitely had to know about SLIP to get your Linux or NeXT boxen to connect to your ISP.
Generational thing. Back when I was supporting banks of USR 28.8s, SLIP and especially PPP were just the job (mostly PPP by that point).
But even then I don’t think I’d have thought to drag a dial-up framing protocol into a new project. Odds are I’d just recreate SLIP from muscle memory.... maybe?
Hahah, I slipped down that rabbit hole about 2 years ago. We needed network access for testing devices on one of our production lines, but the manufacture did not allow wifi at the time. So we ended up tunneling over UART to access internal APIs on the device. It worked fairly well, and was even able to load more complex customer facing react based UIs.
CSLIP is even better.
I wouldn't be surprised if they were trying to remove them from the kernel since you can use a tun interface instead. They would be very good candidates for moving kernel functionality to userspace.
Aren't regular fiber-optics just mass-deployed IP over lasers? (this message is brought to you by at least one, likely multiple, lasers in prod)
Yeah, but kinda cool making it from scratch.
Also this uses air as the medium.
Free-space optical transmission is a thing, though it's environmentally quite challenging. We tested some gear from https://www.koruza.net at the Internet Archive ~10 years ago or so (I've also built some 10G point-to-point links in my garage, though I'd hardly call them reliable). It is pretty cool to see a scratch build rather than using commercial transceivers
Taara is a Google/X moonshot that was recently spun off to do the same thing.
Doesn't Starlink do that already for sat2sat comms?
Not quite the same thing as doing it in an atmosphere.
There's not a huge amount of info I've seen on the specifics of Starlink's LISL setup, but there are a couple interesting bits in here: https://www.pcmag.com/news/starlinks-laser-system-is-beaming...
> Brashears also said Starlink’s laser system was able to connect two satellites over 5,400 kilometers (3,355 miles) apart. The link was so long “it cut down through the atmosphere, all the way down to 30 kilometers above the surface of the Earth,” he said, before the connection broke.
(the presentation that's being reported on, which I don't have access to: https://www.spiedigitallibrary.org/conference-proceedings-of... )
30k would be about 1/3rd MSL air pressure, so that's pretty thin relative to what humanity experiences. Also note it says 'connection broke', not that the connection was way way slower than it would be in a vacuum.
That's right, the beam pointing problem is far harder when everything is moving.
There's this thing called wind that can move your outdoor installations quite a bit ;)
I'd call it a wash, space is hard, but so are atmospheric interactions, weather, foliage, and all the side effects of human habitation (like someone building a house in the middle of your laser link, yes that happens.)
Deterministic space dynamics vs chaotic fluid dynamics (or worse, as the sibling comment by eqvinox illustrates).
Space is hard by many aspects, but on that part it's much easier than on earth.
Space dynamics are not that much deterministic. Gravity itself is kinda noisy (Earth isn't an ideal sphere with uniform density), there's the Moon, orbital decay (caused by drag from particles in low orbit, which is variable), solar radiation pressure (also variable), etc. Calculation of the dynamics will only give an approximate result, a prediction. They need constant measurement of the trajectories and frequent correction maneuvers (by ion engines). But yeah, I think that once the satellites accurately know each others trajectory, their movement shouldn't be a big issue for the lasers, as in the timespan of one laser connection it should be predictable with the required precision. And if both satellites would be on the same orbit, their relative movement should be ~0, so the laser beam ideally wouldn't move at all from the satellite's perspective, the angle would be constant.
Starlink sats do fly just low enough that they experience some mild atmospheric drag. Their next generation sats will fly even lower too. But it's certainly still in the range where simple computer models will be very accurate for at least a few hours out.
Bizarre take.
Atmospheric scintillation is the barrier for free space laser communications on terra-firma; this is one reason we enclose the laser light in optical fibre to avoid this problem.
In space where nobody can hear you scream, scintillation isn't a problem.
About 25-30 years ago we had an FDDI network, and the hosts had optical bypass switches on them.
The FDDI network comprised two fiber rings, one going clockwise and the other anticlockwise. If a host dropped off the network, the optical bypass switch would loop the two rings to each other, creating one big ring. Two non-adjacent hosts dropping off the network would break the ring.
The optical bypass was surprisingly simple. It was a couple of pieces of fiber segment glued to a swiveling magnet; an adjacent electromagnet pull the magnet/fiber assembly, connecting the network rings normally if energized. If power were removed from the electromagnet, a spring would pull it in the other direction, pointing the fibers into loop position, connecting the network rings in looped configuration.
In both cases, the air was the medium between the fibers entering the switch and the straight-through/loopback fiber segments.
Apologies in advance for my poor explanation.
Check out relatively recent Google TPU paper where they used an optical switch to connect neighboring TPUs
Yes and no, worth reading in to. Yes in the sense of, technically laser but gets reconverted along the path in many ways, ethernet turning it in to frames after a receiver (modem, whatever equipment) then translates it.
Laser = over air, susceptible to interference like atmospheric things, dust, flies, also; since it's laser and over a distance, the photons will spread out. Beam divergence.
Fiber lines are carefully engineered to contain the light transmission to get it to where it has to go.
Microwave would be better than laser to my knowledge but then your packets are flying around through the air willy nilly. Things like SSL handshakes and unencrypted hello packets are readable.
But, anything lasers is amazing.
I'd assume most are LEDs or VCSELs. The long range stuff is laser.
The L in VCSEL is LASER.
Oh yeah!
VCSELs are lasers.
The AT tiny is not needed at all. You have a digital signal coming right out of the serial cable which can drive the laser using a buffer and a Schmitt trigger on the receiver.
Eliminating crosstalk is the tough part and requires some modulation to ensure the transceiver isn't accidentally listening to itself via reflections or picking up interference.
Look up point to point laser links. They have been around for quite some time.
Back when 10M ethernet was popular, it was pretty easy to wire up something like this with the transceiver (AUI) port... wire the tx to a laser and the rx to a photodiode (or whatever), ???, profit.
There's lots of commercial equipment in this space too.
When I studied our dorm was connected to the campus network via a laser uplink (and the Internet, with each room having a public IP address). I still remember the foggy daya when people, you normally did not see, gathered in the bars, because the network went unbearably slow. It was nice days with every CS student having a server rack in their closet. I had a VT520 hooked up to it be able to read emails and IRC chats bed without having to wait for my PC to boot.
Has to be the early to mid 1990's?
I was going to say, sounds like my college experience decades later. Still a very congested laser link, and I had a headless minitower instead of a rack (though I wasn't a CS student).
I dreamed of doing something like this growing up so that my neighbor and I could play StarCraft without the latency of dialup. I wish I knew then that it was that straightforward.
We had a laser link to the university 20 years ago. 10MBit/s for 300 people.
That thing was awful.. lol.
The link was dead during
- Heavy rain
- Fog in the early morning
- While snowing for days
- Pigeons building a nest within the optics
There are many long-range WiFi solutions. I wonder how these would have performed as alternatives.
Nowadays you just can buy pair of cheap Mikrotik oder Ubiquiti directional dishes.
Back then this was rocket science.
Google "pringles cantenna" to get an idea of what state-of-the-art long range wi-fi gear was like in 2005...
Yes, these were Yagi antennas.
EBay has many models.
Scale it up, and you have Taara[1]
When I was about 10-12 for a school science fair I built a device called a "photophone" [1] - one of Bell's lesser known inventions.
Basically, a filament flashlight is modified so that a magnetic coil was placed in series. An audio source is then fed through a second coil -- I can't remember the exact details of how this worked. The audio source was one of those fisher price sing-along tape players that I also modified I think. The tape was Abba.
On the other end, a cheapo solar cell was hooked up to a small kit amplifier and then you could hear the audio on a pair of headphones.
This was in 2002 ish, so fibre optics was a thing, but it was basically sci-fi for a scrappy kid in southern Africa. My whole spiel was how this was the precursor to fibre optics, and how one day all communications will be done using light in stead of electrons.
Fun times!
[1] - https://en.wikipedia.org/wiki/Photophone [2] - https://www.youtube.com/watch?v=SxlWrqioifg
Really like seeing networks over whatever laser things. Very cool stuff. This thing always comes to mind. http://images.twibright.com/tns/1208.html
I don't know why, but I would have expected higher speeds. Maybe my mind just assumes "lasers = fast". It would be interesting to know which factors make this setup unreliable at higher speeds.
Most likely one of:
High-speed laser-based systems (gigabit and beyond) don't try to turn the laser on/off at all, they just have something in front that tries to cancel out the signal (e.g., through self-interference) when you want to send a zero.1. Noise in the detection (the room is bright etc.) 2. The laser being a form that's not susceptible to direct modulation (if it takes a millisecond to start/stop lasing, obviously you cannot turn it on/off very fast)Individual bits always go fast, the difficulty is always telling many apart. Speed is kind of the wrong word. It's about throughput
I would bet that the laser pointers used here are the bottleneck.
I have always wanted to give this a try. I have this daft idea of implenting a system logically as an n-dimensional cube and physically on two facing surfaces (even and odd parity processor addresses) with laser connections between the processors. Perhaps it'll be one of my crazy project ideas that I can do in my retirement
I remember trying to my own data transmission over lasers ages ago, back when I didn't really know about how to properly transfer data (PLLs, clock recovery, error correction, etc). It lost timing pretty quickly, but it could just about do a hello world
None of this is remotely new or interesting. I was doing IP-over-laser-light at Ethernet speeds over kilometer distances using free-space optics thirty years ago, and even then it wasn't a new technology.
It's interesting for a hobby tinkering project. Basically like rudimentary transmitters for ham radio. Understand the fundamentals, then you can move on to better quality DIYs
Was that RONJA [1]?
Thank you for your valuable contribution to this conversation.
Isn't fiber broadband exactly this?
It's shooting a laser through a fiber optic cable.
Why is the ping worse than connecting to a remote shack in Mongolia?
I was vibe researching TUN/TAP devices a few days ago. Cool to see such cool projects using them. I was mostly thinking about making a VPN prototype through.
There's a billion different types of VPN.
What problem are you trying to solve?
Increase my understanding
I want to see the actual laser beams while transmitting those packets.
Do not look at the laser, you wont be able to look at it again
I met many people who after grabbing laser pointer first thing they do is pointing it into the eyes of a nearby person, from all the things they could point it to.
I always liked “Do not look at laser with remaining eye”.
A typical computer is connected to the local network using either an Ethernet cable or WiFi. This project connects two computers together through lasers.
No mention of carrier pigeon? IPoAC has three RFCs!
I'd not expect that to be "typical", unless you're in the middle of some pretty huge nature reserve and still need to file taxes online.
Now try IP over electricity.
Powerline is sort of like that. Devices act like an L2 transparent proxies over electrical wiring.
Fun fact: large transmission lines use power line carrier communications for things like remedial action schemes and other system protection functionality (e.g. protective relay trip signals). The carrier can be a few hundred kV, so it tends to be outside the comfort zone of most casual experimenters.
I have UTP and DAC cables here at home. I even have a bit of 10base2 around here, although none of it’s running at the moment.
Photons vs electrons? Just rotate 90 degrees, seem to be swappable more easily than using neutrons or other particles, like tau/muon/electron neutrinos.
Oh wait, didn't fermilab even use neutrinos in 2012? That seems even harder, practically made for an April fools RFC.
IP over Lasers: Fiber optics
IP over electricity: Ethernet
Most IP over Lasers these days is Ethernet (although certainly other standards exist) 802.3ae covers a lot of 10G laser based ethernet over fiber for example.
We know neolithic internet used free air lasers because we haven't found either optical fibre connectors or copper Ethernet in the trenches we dig at neolithic settlements.
Is this a jab at the people who believe Egyptians must have used electric lights because we can't figure out how they lit and ventilated their tombs at the same time during construction?
It's a version of an englishman/irishman/Russian archeologist joke, one finds shards of glass (proves invented optics) one finds copper wire (ethernet) and one finds nothing (wifi, punchline)
Never heard this one
AZIZ LIGHT
Isn't this just fiber optics! Exciting turn of the century tech hitting HN boards in 2025
Every year new engineers and enthusiasts need to be exposed to cutting edge technologies through hands-on introductions like these.