Hacker News Reader

Comment by wiseowise | Hacker News Reader

> A programming language is the single most expensive choice a company makes

By far no. Now I don't know if I even should read beyond that.

This person appears to have been "Languages Product Lead at Google", so they are very used to arguing about how important language choice is.

It CAN be.

If a company chose brainfuck as a main programming language, it's doubtful they'd come back from that choice.

I think the assertion might even have been true for non-pathological cases 20 to 40 years ago. A company that chose Visual Basic or Perl would have had a much harder road than one which chose C# or Python. But i think the languages which have survived to the present day are all pretty close in productivity. Except C.

I've worked for a company with a large code base in Visual Basic .net. Product been in development since the 90's, with rich customer that only cares about their software doing its job. It's a surprisingly productive language combined with Visual Studio. Even though, as a language enthusiast, I barfed a bit now and then. Dev team would like to switch to C# but it would have been a multi-year effort taking away from lucrative feature requests.

that comment is probably referring to VisualBasic 6.0 which was not a dialect of C# like VB.NET but then again your product from the 90s likely started off as that

> A company that chose Visual Basic or Perl would have had a much harder road than one which chose C# or Python.

Based on what? Economics define where you go, not language. If company using Perl or VB has steady cash flow and their bottleneck is language - they’ll just rewrite when it makes sense. No amount of writing in C# or Python from scratch will save you if your product is garbage.

Maybe not the most expensive but it is certainly potentially one of the most costly. Look at Facebook with PHP and Dropbox with Python. It's the sort of thing you can overcome despite the odds (as both those companies did), but can also tip an on-the-edge company into failure.

Same thing with the fundamental architecture of programs, and especially the data model & database solution.

Still I don't entirely agree with the article. He makes it sound like there isn't any difference between programming languages and any preference is purely about developer identity. But that identity doesn't come from nowhere. Rust is popular because a load of C++ developers finally found something technically better. Not because they all woke up one morning and decided to be "a Rust developer".

Where the story falls apart is that Perl is arguably even worse than PHP. One deluded Perl programmer does not prove a principle.

On Rust vs Go, he's absolutely right that Go is has a slightly better "build & deploy" story (though not by a much). But reading between the lines I think he's misrepresenting that - it sounds like a) that was just one point for choosing Rust, and b) that was a point when comparing to typical alternatives, e.g. C++ or Java. It's not untrue that Rust is easy to build & deploy simply because Go is slightly easier.

> Look at Facebook with PHP and Dropbox with Python.

What about them?

They ended up being extremely costly choices. Facebook resorted to developing its own language that only Facebook uses to escape from PHP. Dropbox also tried developing a complete new Python runtime because Python is so slow (but they gave up eventually). Dropbox had to rewrite their sync engine in Rust, and I think they also have rewritten backend stuff in Go.

> They ended up being extremely costly choices.

Extremely costly choice would be choosing .Net or Java and never releasing at all. Creating Hack advanced whole PHP ecosystem and cost nothing in the grand scheme of things.

> Dropbox had to rewrite their sync engine in Rust, and I think they also have rewritten backend stuff in Go.

Bah, idiots, should've just waited 4 years until Go and Rust came into existence.

Reply on news.ycombinator.com