 Hey everyone So is anybody not here for for this session and maybe Move out a bit Or take a seat. All right My name is Alexander suck Working for panto core. We are making a container engine That we hope will be suitable to power any kind of Linux device and better device in particular We are keen on ensuring that whatever we're doing is Helping the full spectrum of embedded devices this includes in particular low-spec devices as well. So we are not talking about a Docker and engine and so on so I wanted to to see if there are like I wanted to meet persons and people that are interested in seeing what are modern topics about low-spec Linux devices these days and And and and see basically if there there's a group of People there that we could basically work with to to figure out the potential challenges that we discover here we had a similar session in in San Diego where We try to Figure out basically what actually a low-spec device is had some back-and-forth discussion on it and then try to To deep dive through it. So I wanted to repeat it here and see What this audience has to say about it? It's more meant to be an interactive session. So it's not just Presenting and listening because we need you input. I think it's about capturing What are is are there any issues these days that low-spec devices need to tackle the Linux community as a whole other software components that are missing? Is there anything that causes particular pain points for you guys and Then maybe see if there is a the way forward to to decide on a gender That we as a community could go together, right? So maybe just to open the discussion. There are a few devices here that first I think we should Maybe assert. What does it mean to be low-spec these days? This is a MIPS based router. I think 16 megabyte storage 64 Memory I assume is that what you guys would define as low-spec device. Is that low-spec these days or is that massive amount of Nice that is something new Oh So there's basically a certain price if you go for for price points There is like a minimum you get and that's the cheapest That's where the minimum is and then if you go lower you have to pay for that other reasons for for doing that I mean power you said Sounds right that was one of the reasons in San Diego people said you would go for low-spec to read try to Reduce the potential to exploit those things That's great to hear that you confirm that story as well But then you might need to pay more for that. I've heard right so this is so security is expensive if you go low I mean alright, so I mean but I mean sounds like this audience nobody assumes that this is low-spec right these days So I guess and then obviously this guy is also Not low-spec, I just want to ensure that people agree This is not low-spec. It looks embedded, but so is this embedded? Right, but I could be embedded But those were the ideas that came up in this San Diego workshop. So I mean maybe we can supplement that I don't know So how can you define low-spec Linux device? Try definition the most cost-effective device you can Get produced these days. Is that what I heard you want to use C++? All right But is there any effective library out there for Linux? I mean a nice standard C++ library that you would use and that gives you still I mean there are like on They're like Python and JavaScript engines and so on for these these so-called microcontrollers these days When like in the last two years or three years with X well, that's lying. No, was it nice? Did it feel good? I Mean is it something reasonable you want to put in some point if there's a hope there's hope, but that's memory That's What's that three megabytes of storage or what what is it I mean there's a tiny linux target, right? So that isn't that I mean I think Michael was talking about that a couple of ELC's bag I mean so when we talk about some do I mean it was always the perception I asked before the session they will said oh we have lots of Waste and people are like misbehaving but then when I looked at the statistics of that it didn't look like it It looked like pretty flat. So people are not adding crafts and Loading up the kernel that much right? Are they really? Where are we now Where are we now? So what is the current size? But if you're talking about memory or that this footprint But there's a lot right I mean I mean all those hope WT currency typically are like two or one five and so on that I see so I mean three is better We know what are they doing there? I mean there's not so many drivers for me And just just without without trying to work against that trend, right? So that's a good data point So is there anything someone should invest in to try to prevent that stuff or do we care or do we not care about these things? like like blow naturally occurring blow because people do not add flags or Disable features and so on So who has a problem to use what So other valid use Linux use cases for the SRAM option stuff for people with suffer and so on I mean I heard yes But not really I Think So you want real runtime tooling to Do that in a lap environment to see what is not needed on the device and then strip out the miner What isn't the hypervisor a bit of conflict with a super low spec thing or Really well, what is it doing? It's hypervisor something right might be right Okay, so runtime LTO Reducing the current size is nice. I Don't know. I mean it most of the discussion has been centered around. I think Linux itself at the moment, so I mean, I but I kind of heard was I mean, maybe it's not worth to really poke the Linux tree that much or invest so much because we're pretty good there and in the end we are only looking for Devices that are the cheapest ones that they have 64 meg of memory. So maybe we say one Way to look at it might be to say, okay, why not say linux should be Just fine to basically serving these lowest cost devices That can mess it here and that have an MMU and so on But with 64 meg of memory, I think the problem is not the kernel anymore, but it would rather be Something else right more user space Topics potentially so yeah, I agree. I mean, I suspect that as well Is it bootloader fine in general? I think so. I mean, I just know that in the open WRT world There was a super Hector bootloader by Qualcomm and so to get super And that was on San Diego the discussion. They said like system D. You cannot use for embedded devices Okay, right, I mean the one argument that they brought up was basically that in system D It's very hard to also be in charge of the boot order and Lots of fiddling right going back and forth and so on So one thing I personally feel is always bad is networking middlewares in Linux. There is nothing My opinion there that is very Comfortable use and does everything we need But there's just my perception that anybody feel like networking is great in Linux for some for low embedded devices But how do you do that? Networking is it Lot of scripts. Mm-hmm. There's a system D network deal. Is that what you're talking about? Does anybody I think Bluetooth isn't that bad, right? I think the blue is blue to blue Z and so on it's not I mean at least when we tried to produce a restrict all version. It was not it was not very Mastiff in size and graphics. I have no idea if there's anybody doing that on Reload spec devices and what to use for that, right? I mean, it's with pure toolkits rendering to framebuff, I guess But then it's not going through the CPU right? It's mostly direct Yes, so what do people think about D bus on the bus side? I think lots of components are using that right said Way to go right So those were the components folks said last time we should potentially take a look at but you guys say system D is great so right, so the distributions that I think the Situation isn't that bad. I thought there's plenty of options if you can afford that kind of space that you see here At this point I'm looking at flash size here not a gram We already in there and we all in this yard of 64 meg rum and then those guys you really have like just they save on storage I know I mean that's what I see those routers right there eight megabyte storage or they have 16 So if you want to do something useful, but you see those in the wild What kind of vertical out there and in which verticals do you see those? Exactly. That's that's one. Let's see the equation that I have in my Exactly and what vertical is that in I mean where do they you said is it for? Controlling machines and stuff or where very part right all right. I Think we are kind of running out of time don't know so What if maybe here from folks So is there anything to do in the loss bag? Segment I had like good ideas from Tim doing like runtime optimizations or memory for the current I Think networking was one of the middlewares that Agreed have might have problems. I haven't really seriously looked at system D network D because I Assumed it's not low enough, but I'm happy to do that be convinced otherwise that this is Providing us with the API said we need But anybody wants to add something But I mean this open W2 you always feeds for me like a it's like a great low spec platform itself But it's also like in very close ecosystem So it's very hard to take procty and just add whatever you want So because nobody is like speaking the language. I think it's you best and so on But I don't think they have a really good middleware to Talk to for networking in particular. I'm not sure if you try to Brought you brought them up because of my networking comment so And just know you trigger them through like configs and you have no All right, all right Okay, I think we can continue discussing at the end after the session whoever wants to do it But otherwise, thanks for the input. I think I got some Good new stuff from the audience here So if you're interested and feel like you would like to contribute or work with some folks on On certain of these areas and feel free reach out to me and have to to connect you guys and Then we talk next time I guess When we in the same same room Thanks