Brian Kernighan on the origins of Unix [LWN.net]

Brian Kernighan on the origins of Unix

Posted Jan 18, 2022 3:40 UTC (Tue) by felixfix (subscriber, #242)
In reply to: Brian Kernighan on the origins of Unix by neilbrown
Parent article: Brian Kernighan on the origins of Unix

My personal hope is for some kind of mesh networking, hardware and software, that has enough bandwidth to make today's centralized networks unnecessary, and has the distributed intelligence to make it practically as fast and responsive as today's centralized networks. The freedom and flexibility of an internet which nobody, google or government, can control, would be a breath of fresh air.

But this depends on more bandwidth and cleverer algorithms, much more on hardware than software, so it's not a candidate for a 9000 lines of code style revolution. I think those days are long gone, just as no one tinkering in a barn can be the next Henry Ford or Wright Bros. It will be something entirely unexpected, spawning an entirely new field.

to post comments

Brian Kernighan on the origins of Unix

Posted Jan 18, 2022 6:16 UTC (Tue) by Cyberax (✭ supporter ✭, #52523) [Link] (1 responses)

> The freedom and flexibility of an internet which nobody, google or government, can control, would be a breath of fresh air.

Google doesn't control networking. Start a web server and off you go.

Brian Kernighan on the origins of Unix

Posted Jan 18, 2022 6:42 UTC (Tue) by felixfix (subscriber, #242) [Link]

It's an idiom. It has a wider meaning.

Brian Kernighan on the origins of Unix

Posted Jan 18, 2022 7:49 UTC (Tue) by PengZheng (subscriber, #108006) [Link] (7 responses)

> The freedom and flexibility of an internet which nobody, google or government, can control, would be a breath of fresh air.

I guess the performance of that network would be terrible. Let's take bitcoin for example.

Brian Kernighan on the origins of Unix

Posted Jan 18, 2022 8:29 UTC (Tue) by felixfix (subscriber, #242) [Link] (6 responses)

Yes, that's my guess too. But I have seen arguments that a mesh network with lots of well-connected nodes might not be, and I don't know enough to judge that. If every node in a city is in contact with 1000 other nodes, is that enough? No current technology that I know of can handle that many simultaneous connections, but maybe it's just a few years down the road and will catch everybody by surprise. Cellular phones took the world by storm in just a few years. Software defined radios exist. Electronically scanned radars revolutionized radar. Who knows what might surprise the world in just a few years?

Brian Kernighan on the origins of Unix

Posted Jan 18, 2022 9:32 UTC (Tue) by Cyberax (✭ supporter ✭, #52523) [Link]

I don't get it. Why would you even do this? What does it achieve, except tickle the fancy of technocrat bros?

Brian Kernighan on the origins of Unix

Posted Jan 18, 2022 11:14 UTC (Tue) by smurf (subscriber, #17840) [Link] (1 responses)

A mesh network requires more power than something centralized. It eats more radio bandwidth, and its latency is by definition higher.

Please tell us why we'd need that, when we already have high-speed low-latency wired and/or wireless options.

(Physical) mesh networks work well for local IoT applications and similar low-bandwidth, noncritical-latency stuff; who cares whether the AC turns off now vs. in two seconds. But try telling a gamer that their network has >50msec to their favorite server without getting laughed off the stage.

Brian Kernighan on the origins of Unix

Posted Jan 18, 2022 11:56 UTC (Tue) by excors (subscriber, #95769) [Link]

And even if it technically worked, I don't think it would provide decentralised control in any meaningful sense. You need an organisation to design the protocol and to update it over the years (to fix security/performance/legal problems with the original design, and to continually improve security/performance/features/etc to keep up with the proprietary competition), and a small enough number of organisations building the hardware and implementing the software so that there's a reasonable chance of interoperability between them all. I don't see any practical way to design a network where users aren't expected to put trust in and give power to someone else.

(See also "web3", which is marketed as decentralised but appears to be already heavily centralised in a handful of service providers: https://moxie.org/2022/01/07/web3-first-impressions.html)

Brian Kernighan on the origins of Unix

Posted Jan 18, 2022 12:18 UTC (Tue) by rbtree (guest, #129790) [Link] (1 responses)

BitTorrent clients handle 1k connections just fine.

Brian Kernighan on the origins of Unix

Posted Jan 18, 2022 12:59 UTC (Tue) by smurf (subscriber, #17840) [Link]

We're talking about a physical mesh network here. Meshing applications is a completely different kettle of fish.

Brian Kernighan on the origins of Unix

Posted Jan 18, 2022 14:56 UTC (Tue) by felixfix (subscriber, #242) [Link]

Good grief. It's just an idea I threw out as something which might happen *in the future* and *by surprise*. You may as well argue against warp drives in Star Trek.

Brian Kernighan on the origins of Unix

Posted Jan 18, 2022 14:06 UTC (Tue) by mtaht (guest, #11087) [Link]

I like what I'm seeing in "tailscale" meshy wize.