first class languages

So like Racket/Scheme and like maybe Ruby?

Factory factory…n is literally just creating an OOP closure for when your language doesn’t support first class functions, closures and/or currying.

Also metaprogramming and abstraction is literally the only way to actually manage and deal with the capriciousness of your stakeholders.

It’s not simple, because it’s literally not that simple. It’s Conway’s Law. That’s what being a programmer in the industry is. I run a platform team, and I get paid because I can organize and deal with technical risk and contingency better than anyone else at my company. You bet your ass I do metaprogramming.

Also my product itself is a factory factory factory. Users create processes to author content, author content, and that content is delivered to other users. All in the same system. Managing complexity is extremely important if you want to work on interesting things.

“And this is the way everyone is doing it now? Everyone is using a general-purpose tool-building factory factory factory now, whenever they need a hammer?”

I’ve had this exact conversation with a programmer who was retiring. He was complaining that I ask too much because I told him that he needed a more generic way to represent the logic that encodes how our end-users traverse the content that our authoring users create and manage. He literally said something to the effect of the above quote to me, but as complaining contempt.

The business explicitly doesn’t want to spend money crafting individual code bases and products and unique logic. Our system lives and dies by our ability to service our internal clients and meet their needs in a dynamic manner. We need manage each factory layer carefully because very often different clients want two different things at two different times, and so each decision needs to be encoded in a way that allows us to make future platform changes without having to sell the business on refactors.

Sure you’ll run into people who overuse things when it could be simpler from the business perspective. But the reality is that most programmers in the industry have never stepped foot into a well run shop. Most programmers in the industry haven’t actually launched a product tip to tail.

It’s very easy to criticize patterns when you don’t actually have to use them, you’ve never seen them being used properly, and you don’t know how and when to implement them.

You don’t know how many times I’ve had to explain what two phase migration means and how to do them across multiple dependency links in the chain.

It’s not about serving assets it’s about hiding telemetry from adblockers, dns filters, ip lists, etc.

Almost every B2C company I’ve worked at, I’ve written or had my devs write proxies for whatever trackers we use. The reality is that every company to whom this data matters to figure out their business model will proxy their trackers. If they don’t they need to fire their lead engineers.

It’s actually pretty easy to disguise this traffic even to the point where you can use the originating server/cdn to interleave the tracking with the content source.

Most companies are incapable of actually building something this techical.

CHIPS was always going to be a giveaway for nothing. It reminds me of that FoxConn CEO’s quote about the fact that the US could pay them to fab in America but nothing would actually come of it. Then they just dropped that $10B Wisconsin project as soon as possible.

That was under Trump. Biden literally saw this and doubled down thinking that TSMC would roll over.

I think one thing you guys should keep in the back pocket, is that Mozilla jobs are the outlier. The average Open Source Developer salary is very close to the US Federal poverty line. They’re paid mostly in comped passes to conventions. Most of the “averages” you see are compiled from data from companies like Mozilla. OSS devs are typically make around $30k in pure cash, even for ones working on large projects. The only OSS devs that make between the $95k and $150k (25th and 75th percentiles) you’ll see online are ones that work for Mozilla, or Intel, or whoever.

What makes this possible is MIT licensing models that corpos shilled in the 2000’s and 2010’s that directly benefit corperate engineering costs, but don’t contribute back nearly the value they extract. If the majority was GPL + copyright assignment, there would be income streams for leveraging OSS projects in closed source applications via licensing deals.

But the genie is out of the bottle on most of these things. See how Amazon is effectively forking an destroying existing OSS models via AWS provisioning of things like redis and elasticache.

After 15 years in the industry, I don’t actually hate cargo cult programming anymore. Cargo cult programing is a useful tool to deal with the industry. Junior devs are going to join a cult, you want them in your cult, and you want your cult to have clear rules. If they want to know why the gods rain cargo, they’ll ask. At one point you don’t have any real control over hiring even as a Lead, EM, etc, because in larger companies saying “no” often doesn’t matter when hiring has been dragging on too long. They need to fill seats for deadlines they decided without you anyway.

As a tech leader with standards, you either need to be in a wonderful company or you need to have a wonderful cult.