favflam

Is this the same as grpc.SharedBufferPool? The gRPC implementation does a lot of memory allocation.

I just bought an Asus Chromebook and it has Debian linux kind of built in. I can apt-get install GUI apps. The employee logs in with their company account (we are a GSuite shop). Vendor-support means there are no driver concerns.

As a company, I would rather employees work in the same environment as the destination of the software.

I can understand Mac workstations and Linux deployment environments, but I can't really understand having Windows workstations with Linux deployment environments. Is it because Windows has better enterprise device management tools?

You get the equivalent setup in ipv6 by having your home modem or router deny inbound connections.

I think port forwarding configuration is a pain that does not offer value over just poking a hole in your firewall to do an authenticated connection over ssh.

You can also check if you have ipv6. I have tried both, but prefer directly connecting home.

I wish coffee shops would start letting wi-fi users have ipv6 addresses. What is holding up the captive portal vendors from allowing ipv6?

What is wrong with protobufs and grpc?

Think of time as defined by how many times a photon bounces between two mirrors. If the two mirrors move at the speed of light with the photon, the photon will never reach the mirror to bounce. Therefore, the photon experiences no time.

There is no universal time T in space time. Time is relative to the observer.

I suggest reading https://en.wikipedia.org/wiki/Minkowski_space for the mathematics.

The Fed cannot fix this.

It is illegal to build dense housing. Car-centric suburbia is fiscally unsustainable and people are feeling the effects via rent. See San Fran and Miami for ground zero.

For what it is worth, I totally agree. I programmed a fairly large project only to find that goroutines were mysteriously blocking in extremely hard-to-debug subtle ways. So, I had to rewrite a huge chunk. I think goroutines are like Ackbar "It's a trap" trap.

I was liberal in usage with goroutines and channels. But after that experience, I decided to religiously track each goroutine in my head. If I reached the point of not being able to mentally map all goroutines running, I would cut out go. I started using callbacks more too to avoid the pernicious blocking I have experienced.

I switched to cycling for transportation. Benefits: 1. way more access to the city 2. eliminated need to go to gym; transport time doubles as cardio time 3. freed up time to see people; 15 minutes door to door is hard to beat

I really hope US cities head in this direction. It is a no brainer for cities: 1. higher density -> lower taxes per person, higher tax rev per sqm 2. lower cost of living -> lower inflation 3. quieter; cars on streets are noisy 4. people see each other more; better social cohesion 5. kids are free from the house to do things

Is PGP out for this use case?

Doesn't the US military have DoD people plug in their ID badges to read/sign emails through outlook?

I read Romans had a height disadvantage against Germans, etc. So Roman strategy was to knock enemies down with the shield and stab up close. Like Mike Tyson vs taller heavy weight boxers.

The shield mitigates the risk of close combat.

40->42->44 C, then drop to 12 C. Then drop the ground dead.

We put developer machines in Kubernetes. And devs ssh in and code remotely. The VS Code remote ssh plugin makes this really painless. And the devs get the benefit of the kubernetes dns to access our internal system they need to program against.

To add a developer machine, we just push a yaml file to a git repo, which automatically spins up the virtual dev machine.

The hold-out resident can just sell the land to a developer who will build something denser than a single family home. Then as a condition of the sale, the resident will own a unit in the new building.

I see it all the time. There is some office building with a weird looking residential unit on the top floor. The resident gets residual rent income together with the developer. New residents take advantage of increased supply of housing and commercial space. It is a win-win-win.

You can theoretically run untrusted code from customers in a secure sandbox. This is a simpler proposition that letting customers run vms.

Nice. Pretty soon, we won't need CDNs for DDOS protection. We just serve API through Onion services.

Is anyone experimenting with the WASI feature?

Question: for sandboxing untrusted code, should I invest time in learning more linux container stuff or switch to learning WASI? I am inclined towards WASI myself.

Walter Rudin's real analysis slaughtered me in school. I can't imagine getting to measure spaces and Lebesgue integrals. Respect to people who get this.

With no street parking, ambulances pass right through.

Also, you don't allow through-traffic of cars.

The only vehicular traffic are taxis (supply limited with licenses), delivery vehicles, emergency vehicles, and those rare people willing to pay $60 to drive into the city to pick up a box or furniture.

I live in a big city like this. Ambulances pass through just fine. The only traffic jams are tolled highways in/out of the city. And those highways are capped at 4 lanes wide with demand moderated with tolls.

If you do a return-on-asset, where asset=sqft of space in a city, you really want to kick out car drivers b/c they are the most space intensive with least return (economic boost/tax revenue). They are like moochers in cafes that buy the cheapest coffee and sit for 5 hours.

Have you considered the costs to contain and clean up Fukushima? The financial costs and loss of lands for decades is a heavy hit for a country to take.

Japan passed a special tax to pay for the disaster (https://www.reuters.com/article/us-japan-fukushima-anniversa... ).

From wikipedia (https://en.wikipedia.org/wiki/Fukushima_disaster_cleanup) >In 2016, Japan's Ministry of Economy, Trade and Industry estimated the total cost of dealing with the Fukushima disaster at ¥21.5 trillion (US$187 billion),...

The Soviet Union, in part, collapsed from Chernobyl.

There is a huge fat tail risk on nuclear power. Roughly and a bit sadistically speaking, coal kills people predictably at a "death-from-car-crashes" rate. There is no fat-tail, you just have a high rate of cancer,etc. It is like letting people driving cars and accepting the 40,000/year deaths from accidents as the price.

Respectfully, suburban guys want to use urban cores rent free. If people with cars actually paid anything close to market rates for the precious real estate they take up with their cars, then most of them would just walk/bike/bus in the city. Car drivers have the government on their side, so they get heavily subsidized real estate (roads, parking).

I don't think this treatment is very equal or fair. Keep your house and your car (SUV?), but pay real tolls and real parking fees.

The US managed to build cities and suburbs fine before car drivers got government priority.

From the "leadership election" snippet, this setup looks perilously close to the peer to peer propagation of transactions and blocks in a blockchain system such as Solana. I suppose the latency requirements are much tighter than 400ms though.

I know Solana does mutual TLS over QUIC authentication between peers and does bandwidth rate limiting via stake. Also, the topology is self organizing. Is it worth copying this tech into a message queue system?

Public transit gets another plus.

Bicycles get a double plus as you do not even tap into a system, nor have a license plate.

The next state is "It works in web assembly runtime (WASI)", no?

I agree. But we should be seeing a spectrum of density with the urban core having some skyscrapers and as you go out from the core, density drops exponentially like a bell curve. Vancover is a bit nuts b/c they have single family homes next to sky scrapers (nimby driven, no?). Janky density distributions are a sign of bad urban planning, endemic to North America.

Side note (slightly rambling): Florida is hitting a wall. Traffic flowed smoothly for a while until the last 10 years in places like South and Central Florida. Miami is full (inefficient urban planning meaning it is prematurely full). Orlando is next to fill up.

Suburbs have a fixed capacity to handle people. Jobs cluster geographically (see Hotelling law[1]), so getting people in low density suburbs to work becomes logistically challenging when the population hits a certain number (~2M?). Like an N to M network problem that expands in complexity exponentially O(n^2). The graph with x-axis in population and y-axis in traffic jams is flat till that limit is reached, then the chart spikes up like a wall.

[1] https://blogs.cornell.edu/info2040/2018/09/17/why-do-competi... and https://en.wikipedia.org/wiki/Hotelling's_law

You can do a POST with an api access token to gitlab.com to get a job to run with custom environmental variable settings. So, you don’t have to use git to start a job.

Yes, but you need a static ip address, resilient server software components, acceptance of occasional inconvenience.

IPV6 works, so long as you have a jump box to get you from a CGNAT'ed IPV4 network into the global IPV6 world.

If you want to skip on the jump box, you could give up some convenience and go with TOR to get yourself back to your home-AWS setup.

I would say this is the biggest barrier to utilizing home compute.

As for reliability, you could use just program stuff to dependency failure rates of 5% (up time of 95%).

I would not run any Docker containers or Kubernetes as that will be too much work. I would stick to bare metal and not bother with isolation. Just make sure you are using certificate authentication or SSH. Do not use a user name and password or expected to be pwned.

Lastly, I would definitely consider using gitlab.com to host code and run a gitlab runner from home. The previous parts of what I wrote become moot. The runner just connects when it can and runs jobs.

I know that Firecracker does not let you bind mount volumes, but QEMU does. So, we changed to QEMU from Firecracker. If you run the workloads in Kubernetes, you just have to change a single value in a yaml file to change the runtime.

I would be scared to let unknown persons use QEMU that bind mounts volumes as that is a huge security risk. Firecracker, I think, was designed from the start to run un-sanitized workloads, hence, no bind mounting.