Adam Stacoviak: You’re just booting into your BIOS, you put your USB stick in there, you’re confirming which – selecting that USB drive to boot from, it boots and you do the install process. It’s no different than the way you would do Linux, or…

Adam Stacoviak: Yeah, let’s speculate this system. What could it actually do? If it ran on this hardware, is it running – is it like a USB version of a software that runs on the USB, but instead it runs on the floppy?

Zac Smith: Yeah, so there’s an organization part of the Linux Foundation called the SSIA, where we created a – think of it like a USB standard for single and two-phase liquid cooling, between server manufacturers, and chip guys like OEMs, and data center operators. So that way we can have the right size coupler, and start to do things like color-coded, so that way if you had a computer, you slid it into a rack, and you would be able to plug a liquid loop into it. The idea here was that if we could standardize that interface, as well as some of the more complicated stuff around fluid flow rates and pressures, that the industry could more holistically adopt liquid cooling, which is dramatically more efficient in terms of removing heat, and using less energy in data centers. And it seems almost obvious, except for computer manufacturers release computers every one to two years, data centers change every 20 to 30 years… They’re on very different timeframes, so there needed to be more of an open standard.

So we’ve been working on that, working along with the OCP crew, as well as different organizations in the Green Software Foundation, which is also in Linux Foundation, to just try and figure out how this standard, which is published under what’s called the JDF, Joint Development Foundation - a little bit different than open source software, which you can actually put in GitHub, and put a BSD license or Apache license on… You can’t do that for just words, drawings on a spec…

And so we worked with the Linux Foundation on the JDF, which has been awesome. And I wish we were more successful, meaning I wish it was more common, but I can see –

Justin Garrison: Well that’s what I’ve been doing. No, this sounds fascinating. I don’t even know you could have like an SD card on one hand, and it’s like connected to the USB on the other side, and you can switch it back and forth.

Adam Jacob: Yeah, it was nuts, right? And it was really condensed. And the Ruby-based Chef server just couldn’t do it, and I remember going to Facebook with the Erlang Chef server on a USB key that had a little – whatever, it was like a little bear’s head on the end of it. And I brought it to Facebook, and we installed it and ran it, and we thought that it didn’t work at all, because the load was flat, and there was no utilization, and then everything finished and it was fine. So the Erlang just literally ate all the load and just didn’t care [unintelligible 01:09:43.18] It was incredible. The first time that we’d used it outside the lab, it was truly incredible.

So we were like “Hey, if you want the good one, pay us for the good one.” And then it becomes hard to maintain both, so then you’re like “Well, maybe we should just open-source the good one too, but then we’ll hold back some features. Maybe it’ll be security features.” Then maybe it’ll be – well, and then what we actually should do is… Like, you can’t hold back basic security. That makes you kind of a jerk. So maybe we’ll build like a dashboard, a better web UI, we’ll monetize that… Maybe it’s security and compliance, we’ll do that, we made some acquisitions, brought in the [unintelligible 01:10:19.12] folks… That was great. They were awesome. Their code was incredible. They helped. We tried every variation of open core possible, and then in the end switched to a model that looked like Red Hat’s, which was the most efficient by a dramatic margin.

Adam Stacoviak: This free USB stick at the front of their lawn, I just – I’ll just run by and steal it. Yet it’s free.

Ian Lopshire: I mean, I think just a regular USB should have worked for this, so I have no idea…

Justin Garrison: Well that’s what I’ve been doing. No, this sounds fascinating. I don’t even know you could have like an SD card on one hand, and it’s like connected to the USB on the other side, and you can switch it back and forth.

Alex Kretzschmar: Yeah. USB connectivity has always been a bit of a risky business.

Justin Garrison: Yeah, no developers on call… What even was – on call, you’d have like a USB [unintelligible 00:36:44.28]

Kris Moore: [08:18] Yeah. If you went and set that up, then you should have mirrored redundancy. We see less and less people doing that these days, but in the early days of FreeNAS in particular, that was a lifesaver… Because a lot of people were using cheap boot media, USB sticks in a lot of cases, and those are not great, as you can imagine… You know, that $20 or $10 USB stick you has at the local Walmart probably won’t last that long.

Adam Stacoviak: [00:51:55.16] Yeah. Well, I was actually thinking about this recently, and I just did this with screenshots. I did a fresh install of – because I’ve been messing with Ubuntu 22.04, or sorry, 23.04… And I got a redundant OS installation, I’ve got two discs, I’ve got a swap, I’ve got a boot, I’ve got root, and all that stuff like that… And so rather than just choosing one drive, I want to have the system be fully redundant by having two drives in mirror. And I like to show that off, either in written, but the only way I could do it really was like through screenshots, and then [unintelligible 00:52:27.25] those screenshots. Now, will I do a full emulation? It’d be kind of cool to have all that I already have, but then at the end, or somewhere else, a sidecar would be like “Here’s literally the environment to go and do just that. You’ve got two discs, so when you get to that part, you can configure these discs, because you can follow my instructions.” So rather than having to pull down a VM, or Proxmox, or actual [unintelligible 00:52:50.09] you take a USB stick and boot up into it and do the full thing yourself… It’s accessibility to what’s kind of trivial to some, redundant OS installation on Linux, but there’s a lot of steps in there. There’s a lot of steps in there in like choosing the partition, adding the partitions and giving them the paths, and stuff like that, and adding them… It’s a mess, really. So I want to do the example through screenshots, but the best version of that really would be an interactive playground they could do… I mean, just follow the steps.

Steve Francis: Yeah, I was trying to think about number twos, and like most of the number twos are things that the Image Factory is solving; it’s cumbersome, and if you have to write a system extension, and how you deploy that, it’s like, that’s not a good user experience. How to enable GPUs wasn’t a good user experience. With Image Factory, that’s like oh, now “I want this, with this extension”, and the USB drivers, and the GPU support…

Breakmaster Cylinder: I disappear into the desert for 40 days, I come back with music… I come back with a USB drive [unintelligible 00:19:37.29] Is that what you wanted?

Kris Brandow: USB 4.0 is basically Thunderbolt 3.0, so it can put PCIe over it. So yeah, USB 4.0 and all of that… This section of things is also like a deep rabbit hole that I also want to talk about more, because I think it’s fascinating how much we cling to abstractions that don’t make sense anymore.

Adam Stacoviak: Okay. So if somebody out there wants to learn how to support USB controllers on a Raspberry Pi, they can go to SerenityOS as an outlet to enable, to be the proving ground to hack on, right?

Jonathan Carter: Yeah, the installer used to detect that you have a Wi-Fi card that needs firmware, or a sound card that needs firmware, and then it would prompt you that, you know, if you have a USB disk that contains this, you can insert that now and it will load that firmware to continue to install it. Now that firmware isn’t included on all the installation media by default… So at least it doesn’t need to do that anymore. From your perspective, the hardware will just work.

Adam Stacoviak: What I like about it is that it can run Plex, and I don’t always want to phone home for Plex, because the network isn’t always there, really. And so if I can actually have one – the reason why I like it is for all the reasons you’ve said; it can be the router for all the devices that I trust, I can connect to LTE, or a hard line to the local network if there is one available. If it’s an Airbnb, or this RV place happens to support that. Sometimes they’ll give you a router that connects to their network, which you can then run an Ethernet port out of to something else. So you can still be their device, but then protect yourself from there. But the fact that I can run Plex and just run a mobile version of my Plex, so it doesn’t have to be my main version of it. I can just pull over the kids’ directory, for example, which is - they have their own directory. All the stuff that the kids can watch is in the kids directory. I can just clone that to this device. Now, obviously, if it’s too big, and it’s 10 terabytes or whatever the number might be, maybe I’m then choosy, or it’s a selection process. But I can always take a mobile version of my Plex, essentially.

And I wanted it for all the reasons - it’s low-power, Intel CPU, virtualization, can run Proxmox… So thank you for doing all the hard work, Tim. And then the relationship, I’m sure, because – finding hardware manufacturers that support open source the way that Protectli seems to based upon what you’ve shared about them - it’s kind of hard. If it weren’t for folks like you, and I think just YouTube in general, or like content creators, either being approached by these brands, or being curious and having these needs, and being like “There’s got to be a better way.” If it weren’t for folks like you going and doing all that digging, then people like me would not be able to just piggyback off all your work and just – because I’m gonna go and figure out how to do that for myself, because my idea was “Okay, I’m gonna get that small square USG, a POE thing with a POE access point…” I was gonna do a small version of UniFi network. And like that thing basically is all that in one, plus it can run Plex. And also have – I think it supports SATA SSD, so I can maybe doing an eight-gig, or maybe I can plug in… I think there’s some USB ports, or something like that, too. Maybe I can have an external device that’s just for the storage. I don’t know, I haven’t thought that far, because I’ve only recently just watched your video. But that to me is like, that’s the better way to do it.

Steve Yegge: And it’ll be polished, right? I mean, look at your journey on this show, from Skype and mailing USB sticks around from your guests, to Riverside today, right? Stuff gets better. And a lot of engineers are young, and they don’t see that the world used to not have AWS and Kubernetes. They don’t see that – and that makes it harder for them to suspend their disbelief. I love your framing, Adam… Because, right now, if you look at it, you’d be like “Well, it’s wrong sometimes.” That was a huge stumbling block for IDEs in the early days. A lot of Java programmers wouldn’t – any strongly-typed language wouldn’t take an IDE, because they weren’t accurate all the time, until they finally evolved to the point where they were… And during that period, there was a big war between the command line folks and the IDE folks, for 5-10 years, before finally the IDs were good enough, and everybody was using them. Coding assistance is exactly the same. It’s early days, we’re basically getting the early adopters, bleeding edge, but it’s going to improve in quality and refinement until it’s pretty obvious that you have to use it, or else you’re behind.

Mat Ryer: Yeah, so it’d be USB. Wired or Bluetooth?

Alex Russell: Sure. So my name is Alex Russell. I’m a product manager, I guess, now, on the Edge team at Microsoft, and I’ve been working on browsers for 14 years, something like that. I joined the Chrome team before 1.0. But prior to that, I’d spent something like a decade building extremely large JavaScript applications, often for folks who had real problems. So I was the lead for the Dojo toolkit project, which was the conglomeration of a bunch of other sort of proto DHTML libraries back in the early 2000s… And that became kind of the corporate standard for IBM. We used those tools to build things like business intelligence dashboards, Ajax business intelligence dashboards for folks like Informatica; we built entire PIM suites out of it, helped Sun build stuff… I think some of those systems are still running. For a bunch of years, they were – the AOL email client was based on our technology, calendaring, chat, email, the whole thing.

So around the same time, a lot of my friends in the early kind of proto JavaScript days had been getting hired by Google, and eventually they pulled me in, and so then I spent the next 14 years or so working on browsers, but specifically the web platform. So I helped lead projects that developed various aspects of ES6, things like arrow functions and classes, promises, async and await… We developed at the same time what became web components… We’ve prototyped a bunch of new things in CSS that unfortunately had to wait a much longer time to make it into the world… And then from there, develop PWA’s and service workers, and helped lead the team that did push notifications, and then we did project Fugu, where we’ve been expanding the power of the platform to do things like serial, and HID, and USB, and Bluetooth, and NFC, and access to low-level camera stuff, web transports, web codecs… Oh, and I spent about a decade on TC39 in the middle there, and then ran standards for the Chrome team.

So these days, I spend about half my time working directly with Microsoft’s largest web products. And before that, since 2015 or so, I’ve spent a ton of time doing this kind of incidental consulting with people who are building PWAs. Not because I want to care about performance. In fact, I actually kind of don’t want to think about it; it’s just that whenever we sit down to work with teams, trying to figure out what hurts, that’s always at the very front of the list, whether or not the team knows it. And that’s been extremely depressing, kind of Groundhog Day thing, since about 2015.

I remember we’d sit down before every Google I/O, or for every Chrome Dev Summit, with a set of teams that we wanted to highlight their incredible work on a product, working with these incredibly smart engineers, and they will have built it out of the new hotness in JavaScript land, and it will perform like absolute dog***t. Just not okay. Just absolutely not okay. And so then it turns into this remediation process.

[08:07] These are - and I cannot stress this enough; these are not bad engineers. These are people whose houses that they’ve constructed are extremely flammable, not because they didn’t know how to join the corners, and put up joists, but because the materials that they were sold were not fit for purpose. So a lot of things ended up in their hands…

And so then we would sit down… And a lot of this work turned into sort of more incidental consulting. And again, I can’t stress enough - performance is just the first thing on the way to getting to where we really wanted to go, which is to say, “Okay, have you tried out these cool, new APIs we’re trying to make the web more powerful with? We want to like help you learn about what you need, and then identify the next set of problems.” Because that kind of consulting, that collaboration is how – if you’re a platform maker, you can go around and like you could give everyone a survey and say, “What do you need?”, and you’ll probably get a bunch of answers that are just like “Please make me a faster horse.” And the only way that you really learn about what people actually need for their products is to go sit with them.

So it was an attempt, and it’s a continued attempt to try to learn directly from people building products about what’s hurting most, so that we can go and like add those things to the platform. But this performance thing is just this constant irritant. It’s an ankle-biter at every moment. We just can’t seem to get past it… So here we are. Yet another team at work was extremely let down by a bunch of tools that have been sold to them on scalability, and the only way that those tools scaled was directly downwards, like a lead balloon. And it’s not the team’s fault. They are extremely strong engineers; extremely strong engineers. I can’t stress this enough. I work with great people. I’ve been lucky enough to work with like more than 100 organizations over the last eight years… They’re filled to the gills with good engineers who have been absolutely let down.

Steve Francis: Yeah. Well, so Talos does make setting up a Kubernetes cluster really simple. Omni makes it next-level simple, where it’s really – so Omni is our SaaS service for the installation of Kubernetes. So the way it works is you log into your Omni account, your SaaS portal, you download the installation media for a Raspberry Pi, or an ISO for a bare metal, that you can put onto a USB or whatever, or an AMI for Amazon, or Google, or Oracle, or wherever you want to run your compute resources… So basically, you boot your machine wherever it is off that image, and that’s basically it. So that image that you’ve downloaded has built-in WireGuard endpoints, and joint tokens, so as soon as the machine boots, it registers with your Omni account, and it shows up as an unallocated machine in your web portal.

And then, all you do is you go into your web portal, if you want to make a new cluster, you go to your unallocated machines and say “This one’s a control plane. This one’s a control plane. This one’s a control plane. Worker, worker, worker, worker, worker. Go. Create cluster”, and that’s it.

And WireGuard gets deployed, KubeSpan is configured, Talos is installed, Kubernetes is installed, the cluster is bootstrapped… But that all happens automatically; you get a nice management GUI where you can see performance, and nodes, you can run upgrades… It’s really simple, really slick.

Jay LaCroix: Well, first of all, I would say the biggest problem with your way of going about is that you should be replaying Castlevania, Symphony of the Night from the beginning, at least twice a year. And if you’re not, you’re doing it wrong. [laughter] Because it’s one of the best games ever made. But I mean, that’s one of those I tried to beat at least once a year. But to answer your question, though - you can, it’s just going to be very time consuming. And Castlevania, Symphony of the Night is going to be one of those that you kind of have to beat it more than once to really appreciate it, because there’s going to be unlockables, and things…

There’s a memory card adapter for the PS3, but you’d have to have a working PS3 to use it, which is the hard part. I’m not aware of a way to just grab it off the hard drive. But if you could get it to boot and then source that memory card adapter, you can get a PS 1 or PS 2 memory card, and - I don’t know if it has to be one or the other, but you could basically copy that save file from the Playstation’s menu to a physical memory card, and then use a memory card or USB adapter to dump it to your system, your PC, or whatever.

It’s going to be a little bit of work… And the memory card adapter was – at least it was rare. And I know it’s very collectible, so you might be actually needing to prepare to spend some money on this; and it may not be cheap.

Nick Nisi: I use those all the time, but it’s a constant battle of, “Oh, I want to vertically center this, and I put the wrong one in, and then I put the right one in…” Or I put the right one in, and then the wrong one in, and then the right one. It’s like a USB, it’s like some quantum state, I have to be in three states before it actually works… But all that to say I can never remember which one does which. [laughs]