mborch

While ECR costs a lot less than DigitalOcean's container registry service, they still charge 5x more over AWS S3. I suppose if you need encryption? Seems like they provide some value-add that some users might need.

This project was mostly just a they said it couldn't be done kind of thing.

As I said, he brings out some compelling arguments. Making a living based on your research work is not in itself a cause for concern. That's what most researchers do.

There are some compelling arguments brought against this idea here: https://ianmsc.substack.com/p/no-masks-do-not-work-against-t....

That's not the complete story on these past two years and the flu of course, but according to the narrative presented in that article, we don't have much evidence to suggest that masks had anything to do with it.

The problem is that many projects (open or closed source) don't respect the exponential cost in bringing in a new dependency, i.e. code that's not under your own management.

Bringing in a shitty old bloated logging library with functionality you don't need or want compounds the interest even more.

Should be possible with "SET CONSTRAINTS".

That's probably more or less the same thing in terms of what actually happens in PostgreSQL.

All things considered, an actual database probably gives you the least gray hair, but with some careful test setup I have had good success using the savepoint/rollback trick (and it trivially supports nested fixtures as well).

Another trick is to use (possibly nested) savepoints and rollback after each test, never actually committing any data.

When I look at the headlines you posted I don't see anything that remotely leads me to think this is related to such media as you suggest.

The article suggests that the evidence decision-makers now look at when it comes to masking has shifted to lesser-quality observational studies and makes the argument that this has clouded the ability to make good decisions, for example not prioritizing other strategies such as securing better ventilation.

He argues that it's basically just poor decision-making as a result of "politicized science", politicians need to be visible to the public and what's more visible than a mask mandate (or closing down borders).

This list seems to be another example of a mediocre programmer trying to look better by saying that an actually better programmer does not exist because their output is objectively not better when you look at it holistically.

The reasoning seems to be that this better programmer would write programs that are impossible to understand or maintain or a host of other reasons why the person is basically incompatible with a long-term sustainable software output.

There are other reactions to being a mediocre programmer, for example trying to get better or simply giving up entirely.

Better programmers make mistakes too, sometimes big mistakes. But they also break new frontiers.

Same audience as https://julialang.org/ it would seem.

That masking wearing is virtue signaling, showing that you care. Which is not necessarily a bad thing, but you might as well wear a t-shirt with a message.

"Very few teams are actually practicing CI/CD" is mentioned in the article.

In an enterprise setting, due to segregation of duties, it's often true that CD is impossible due to restrictive change management processes that result in unpredictable and typically week-long cycles.

Life expectancy + 10 years, not sure what the story is here. People don't live infinitely.

From the docs:

> We rarely recommend running with more than 5000 threads. If you seem to need more than 5000 threads, it’s very likely that there is something not quite right about your setup, and you should investigate elsewhere before you increase the maximum value.

Why? It's a serious hate crime, akin to instilling violence.

Hey HN. While the focus of this driver is not performance as such, a comparison can be found here:

https://github.com/MagicStack/pgbench/pull/2

What's more interesting for fans of type safety is that the driver uses PostgreSQL' extended query protocol to correctly process and encode query parameters to match the required type specification as returned by the database for a given query.

It uses binary mode only which also makes the codebase smaller and provides a significant increase in performance.

What it does have is compile-time parameters where you can bind a type variable during compilation:

  fn max(comptime T: type, a: T, b: T) T {
      if (T == bool) {
          return a or b;
      } else if (a > b) {
          return a;
      } else {
          return b;
      }
  }

The type variable isn't used at run-time.

It's not just software that doesn't "square the very definition of a project ... with our real world experiences". The same happens on the business side. There's always the other side of the equation: how the system reacts to what you're trying to do.

Perhaps some component doesn't really gel the way you had hoped for. The same goes for business opportunities. Sometimes it just doesn't work out and it can be a real blow to the project. And sometimes the users just aren't there or don't appreciate what you've done.

Management generally has to be flexible, not just in software, but in a lot of divisions. Because ... unknowns ... that are discovered as you go.

He wants to make that case that software development is a more pedestrian affair than it might seem, without much of an argument. But it's not true.

Even a small shop needs to live up to pretty much a global standard, not least in terms of security. In the EU you must engineer with a security perspective in mind. The field is constantly moving. People come and go faster than most other industries.

I think there's an analogy to getting a better bike: it doesn't get easier, you just go faster.

From Google Analytics.

That would be illegal (in the EU, at least). Why would a successful business risk that?

or better, just spend more time outside.

But they are... Unless you want to be pedantic.

It most certainly isn't. The Wikipedia article has an accurate description. Google pulls a wrong definition for some reason.

The brain can drive a car and uses a fraction of that amount of energy.

The new web needs to be distributed in a privacy-preserving sense. Today, you can't realistically browse the web without getting identified and – generally speaking – geolocated.

What we need is a model where you pull information you request from distributed and diverse pools of public domain content.

For most people that's currently Google due to search-based location bar.

Cisco's got a royalty-free codec that would let you decode like it's 2017. https://github.com/cisco/thor

It uses Amazon Redshift under the hood.