[collyw]

> There is obviously a large problem with the way we do software engineering, and we are still figuring it out.

> The longer you spend in software, you realize that the "accepted way to do things" changes all the time.

Why though?

We had it figured out reasonably well around 2010-215 in my opinion. Then we got more fads, NoSQL, frontend frameworks, cloud infrastucture, GraphQl in place of REST. These aren't required in the majority of places where I have seen them used, a standard old school lamp stack would do fine.

Everything is more complex, takes longer and solves imaginary problems that most companies don't actually have in my experience. Sure there are a few that will need that scale, but no one aims to keep things simple any more.

[AA-BA-94-2A-56]

A lot of newer technologies solve problems that only very large companies face— mostly to do with scaling multiple systems and synchronising their state.

It’s in these companys’ best interest to recommend these technologies to all developers and companies, because it increases their pool of available hiring talent.

That’s how you get startups writing from the ground up, a micro-service backed SPA with layers of abstraction for the data to pass through; when honestly, a SSR monolith backed by MySQL would be better for them.

[valty]

GraphQL is a great example. Try stepping through all the first and third party code involved in serving up a simple list of records from an SQL database. Mounds of useless abstraction and boilerplate. YAGNI. But the same could be said for SQL databases...