rlyshw

This is kinda getting at a core question of epistemology. I’ve been working on an epistemological engine by which LLMs would interact with a large knowledge graph and be able to identify “gaps” or infer new discoveries. Crucial to this workflow is a method for feedback of real world data. The engine could produce endless hypotheses but they’re just noise without some real world validation metric.

Cool. I've actually been playing with QEMU internals a lot recently. Specifically with the multi-process experimental features. Although I can't seem to find any consistency on where the main project is headed. They admit that the documentation pages can be well out-of-date with the upstream implementations, but they seem split-brained even within the code.

The main project ships with the multi-process qemu approach, mostly defined in their docs: https://www.qemu.org/docs/master/system/multi-process.html https://www.qemu.org/docs/master/devel/multi-process.html

But I saw some update buried in a mailing list that development of the multi-process system has been superseded by vfio-user, mostly led by nutanix: https://github.com/nutanix/libvfio-user

The nutanix repo refers to an oracle-led fork of qemu with the full vfio-user implementation built-in: https://github.com/oracle/qemu

So, they're still separate projects, right? Well, kinda. the mainline project has the vfio-user-server implementation merged in: https://github.com/qemu/qemu/blob/ac5f7bf8e208cd7893dbb1a952...

But not the client side (vfio-user-pci). So, the feature is half-baked in the mainline project.

I don't know if any of the qemu devs browse HN but it would be nice to hear more about the plans for vfio-user.

I’ve thought about doing this, but a few reservations came up when I considered getting started with a family friend. I just pictured a contracted ”IT MANAGER” getting rabbit holed into some time-sink extreme;

1. Dedicated operational IT admin: Dealing with repetitive tasks+requests, like managing customer’s Microsoft environment and on-site infrastructure.. Owning physical and AD infra doesn’t sound like a part-time job.

For e.g; a/v and physical IT asks; like conference room operation maintenance and support, Desktop workstation triage (have you tried turning the monitor on?). The dreaded “can you set up the printer?”…

And what if the customer sets me up as their site’s dedicated AD domain admin? Resulting in repetitive requests for user/access management CRUD operations. And/or micromanagement of tedious things like email and mailing lists…

Or

2. Dedicated software developer, website or business workflows.

Building a website and getting micromanaged or overburdened. (“can you change the logo to blue?” “Can you redesign the whole home page?”)

Or, get pulled deep into providing a business-critical software workflow or application. Fielding sales/exec requests, interpreting their business requirements, and then building AND delivering (for e.g a customer management system) is not a part time job…

How do you operate to keep the scope limited? What steps help buffer yourself from a slippery slope of full-time services?

Helm charts are pretty standard, git/kube-native, and lead much more nicely into the more sophisticated operator/controller model.

One part I haven't quite put a handle on is where/when to use CRDs in either helm or OLM. The community still seems split on how to use CRDs.

USPS might be mired in fed scale problems. Maybe a Library is more appropriate? At least, more directly accessible at the local level. I’m just not sure how exactly that would work, or operate thru existing library organization…

I think the incentive of a trade/artisan economy would make more sense, and justify individualized labor (house calls for NAS reconfiguration, for instance). Like a plumbing contractor vs inspector… I like the socialized idea, but I don’t see how the implementation would work under current social service labor system and organization…

I know that this is a typical HN post, assuming everyone should become a Linux sysadmin. But related to the parent, and recent developments in Zero Trust Access products, I wonder if there is a pathway towards neighborhood-scale sysadmin services.

I mean, I essentially provide that to my small social community with a private media tenant.

With ZTA systems in place to accommodate remote access, maybe there is an appetite for neighbor-to-neighbor network sysadmin services? Hard to compete with the sleek silos of big box brands and their infinite marketing budget, plus 5 9s of service, though.

Linux has been pretty stable for decades now. I’ve been using the same core configs and bulk data in my home compute environment basically since I started using Linux. Remote repos for any syncing needs, then just tar/rsync bulk archive data over. Store longer term or stale data on older decommissioned HDDs.

I’ve been running more or less the same services through hardware, hypervisor, and now kubernetes migrations and revisions. It seems to me doing things “the Linux way”, sticking to open source where possible, is resistant to the fast pace of the consumer innovation market. When anything new comes along, it’s usually relatively trivial to transfer over.

Seems like a lot of ongoing o&m and sustainment of an existing product, similar to any other day-to-day.

Do you have insight on what the proposal, design review, and first commits look like when architecting a brand new project? Like what do the first 30-60-90 days look like from product idea up to those o&m and sustainment activities?

Im building with featherjs[0] right now and I love it. Jwt, user handling, routing, and (most notably to me) real-time functionality is all built in. Probably the most rails-like backend framework I’ve worked with in Node so far.

[0] https://feathersjs.com/

Except the Spotify model is also rife with issues. Artists generally hate Spotify and hardly make a living off of “pay per stream”. Most of them still very much depend on tours, merch, and, at the higher level, brand deals to make any money off of their craft.

I recently had an issue where my UDP service worked fine exposed directly as a NodePort type, but not through an nginx UDP ingress. I _think_ the issue was that the ingress controller forwarding operation was just too slow for the service's needs, but I had no way of really knowing.

Now if I had this kernel level network monitoring system, I probably could have had a clearer picture as to what is going on.

Really one of the hardest problems I've had with learning/deploying in k8s is trying to trace down the multiple levels of networking, from external TLS termination to LoadBalancers, through ingress controllers, all the way down to application-level networking, I've found more often than not the easiest path is to just get rid of those layers of complexity completely.

In the end I just exposed my server on NodePort, forwarded my NAT to it, and called it done. But it sounds like something like ContainIQ can really add to a k8s admin's toolset for troubleshooting these complex network issues. I also agree with other comments here that a limited, personal-use/community tier would be great for wider adoption and home-lab users like me :)

Ha, this is great. I just spent the holiday break launching a kubernetes cluster in my home lab, on which I’ve deployed a Minecraft server to play with my friends. I guess with this I could manage my Minecraft kubernetes deployment WITHIN the world itself. One creeper and the whole server comes crumbling down. It would be a very risky game.

Love this. A great point. HLS via CDN is really just "downloading files but the source is provided kinda fast"

Yeah as the sibling comment mentions these WebRTC implementations do not scale. While you "can hook it up" for hyper-specific applications and use cases, it does not scale to say an enterprise, where a single SA needs to support LL streaming out to tens of thousands of users.

I imagine the (proprietary) stadia implementation is highly tuned to that specific implementation, with tons of control over the video source (cloud GPUs) literally all the way down to the user's browser(modern chrome implementations). Plus their scale likely isn't in the tens of thousands from a single origin. Even still, I continue to be blown away by the production latency numbers achieved by game streaming services.

And my use-case is no use-case or every use-case. I'm just a lowly engineer that has seen this gap in the industry.

A quick search for "latency" in here has one little hand-wavey blurb about Mux working to optimize HLS.

>Using various content delivery networks, Mux is driving HTTP Live Streaming (HLS) latency down to the lowest levels possible levels, and partnering with the best services at every mile of delivery is crucial in supporting this continued goal.

In my experience, HLS and even LLHLS are a nightmare for latency. I jokingly call it "High Latency Streaming", since it seems very hard to (reliably) obtain glass-to-glass latency in the LL range (under 4 seconds). Usually Latency with cloud streaming gets to at least 30+s.

I've dabbled with implementing WebRTC solutions to obtain Ultra Low Latency (<1s) delivery but that is even more complicated and fragmented with all of the browsers vying for standardization. The solution I've cooked up in the lab with mediasoup requires an FFMPEG shim to convert from MPEGTS/h264 via UDP/SRT to MKV/YP9 via RTP, which of course drives up the latency. Mediasoup has a ton of opinionated quirks for RTP ingest too, of course. Still I've been able to prove out 400ms "glass-to-glass" which has been fun.

I wonder if Mux or really anyone has intentions to deliver scalable, on cloud or on prem solutions to fill the web-native LL/Ultra LL void left by the death of flash. I'm aware of some niche solutions like Softvelum's nimble streamer, but I hate their business model and I don't know anything about their scalability.

I'll have to look into it! Now that I've got the 5G-UW plan from Verizon, I wonder if I could actually get >1Gbps to my laptop via wired hotspot.

Sorta related;

I often travel for work and was having an annoying time with Verizon’s hotspot throttling. 4G LTE should be able to run at ~5Mbps but devices connected to the hotspot get throttled down to an insufferable 400Kbps. This was super annoying and felt like an arbitrary action on Verizon’s part. I found out I could run a SOCKs proxy on my iPhone via a Pythonista script and tunnel hotspot connections through it to fool Verizon’s throttling systems. Worked great, even though the UX of launching a Pythonista script as a service and pointing clients to it was slightly clunky.

I read the “what exactly does this do?” section and still don’t know what it does. Maybe I’m just dumb, but there doesn’t seem to be any “user story” to succinctly describe what the heck this accomplishes... I just use ctrl-c on windows and cmd-c on Mac, it’s not that hard. But maybe this is way bigger than just... remapping win-c to ctrl-c? Idk? There’s no story about this utility directly in the Readme...

I assume you are getting downvoted for not adding context so I’ll help; urbit is literally designed around the principle of total ownership. The community (purposefully) does a terrible job of explaining it because of some enlightenment complex but the promo video put out by Tlon does a pretty good job of summarizing. https://youtu.be/M04AKTCDavc

I believe urbit is the solution, just waiting for the implementation to get polished up.

I've done installs of video streaming tech on USAF-managed NIPR net and we've had to reverse engineer some of the mechanisms on their own network... This does not surprise me at all.

the transparent content suggestion algorithm is insanely good. If you spend any amount of time in the actual app, you'll find it immediately starts catering content to you based on your specific interests (without ever obviously indicating them). Pair that with a very simple yet powerful "editing" suite, complete with audio editing and vfx, the app is the perfect storm for micro video content creation/consumption.

Once you get the infrastructure set up and cloud-init-enabled template images, servers can be very turnkey. I can spin up a new DNS-addressable, auto-updating Ubuntu VM in two or three clicks then deploy software like this on it from the CLI. There’s even a system for easy LXC container apps called turnkey Linux. Works great with Proxmox.

The spatial audio feature makes me think that Apple has some level of head movement tracked with airpods.

I'm imagining a system where you can just use nods/head shakes to move through some sort of binary decision tree to execute some basic interactions, like reacting to incoming alerts/messages.

new text, music pauses, siri reads it to you: "Your mother asks if you'll be home by dinner, Would you like to respond?" shake head no -> interaction cancelled, music resumes nod head yes -> "Ok, how would you like me to respond? Yes, you will, or no, you won't?" gesture head for appropriate response

Easy? Dumb/ridiculous? Sure, you can't get suuuuper deep with the decision tree and it's tough for non-binary responses, but it's enough of an interface to have a meaningful, non-verbal engagement with a computer.

What about the WebRTC part?

The post ends at RTP out from FFMPEG. Maybe I’m supposed to know how to consume that with WebRTC but in my investigation it’s not at all straightforward... the WebRTC consumer needs to become aware of the stream through a whole complicated signaling and negotiation process. How is that handled after the FFMPEG RTP stream is produced?

can't even access my vanguard account. Hearing reports from friends/family across the board that they can't access their financial accounts. Fidelity, Schwab, more.

In my experience, something like 90% of content on Instagram Reels is not originally created there. The recommendation engine (accessed via the search page) is almost exclusively bot accounts that steal and repost low-effort attention-grabbing content from other socials. Even the original content direct from creators usually has the TikTok watermark on it since it was originally created and posted there. In my experience as a casual Instagram surfer, it is wholly terrible compared to tiktok.

This is my same exact thought. Although sometimes I think the blinding light might be a good disincentive to stare at my phone in bed.

I don’t like having “one-off” apps on my phone for something that safari is more than capable of handling. I prefer to use the web versions of Twitter, FB, and Reddit since they all work fine as just standard modern web apps right in the native browser. (Although I’m rarely on Reddit anymore due to the app-first nag campaign...)

Not sure about the OP but I’m often browsing on iOS where this isn’t an option.

This is incredibly cute. Super satisfying to see all of that antique gear in such good shape. It's interesting to think about the progression of various computing interfaces, and compare it to some of the cutting-edge AI-based interactions that we are seeing emerge today like GPT-3.