Hacker News Reader

Comment by p-e-w | Hacker News Reader

> Software is still the same mess it was when I started writing and working, or perhaps even worse.

I disagree pretty strongly with this claim. Nine years ago we didn't have (or they were in their infancy): Rust, Go, TypeScript, GitLab, VSCode, Language Servers, American Fuzzy Lop, Linux distributions that work out of the box on most hardware, web "standards" that are actually standards, ...

I would argue that all of these have dramatically improved software quality, and the software ecosystem as a whole. The industry is unrecognizable compared to a decade ago, and almost all changes have been for the better.

View on HN

> I disagree pretty strongly with this claim. Nine years ago we didn't have (or they were in their infancy): Rust, Go, TypeScript, GitLab, VSCode, Language Servers, American Fuzzy Lop, Linux distributions that work out of the box on most hardware, web "standards" that are actually standards, ...

Of those I think the only one that he uses/likes is Go — although I don't know what "American Fuzzy Lop" is and I'm uncertain about LSP. He uses old hardware, intentionally unsafe languages, probably basic terminal editing, and dislikes modern web such that standardizing it is only standardizing nonsense. He even has or is deeply involved in more 'traditionsl' alternatives to many of these — namely Hare (rust/go/etc), Sourcehut (Gitlab), and Gopher (web).

From his perspective, the changes to software that have occured are merely incremental changes to an existing pile of garbage

While I look at it and see incrementalism. And churn.

Yes, fuzzers now are better than they were the first time I fuzzed a system in the late 1990s.

Yes, VSCode is better than the Turbo Pascal environment I used in the 1980s.

Yes, GitLab is better than rsh/ssh-ing into a centralized CVS repo like I did in the 1990s.

Yes, CI systems are miles ahead of the manual "daily heartbeat" builds people did in the 1990s.

But "unrecognizable compared to a decade ago"? Maybe in some niche areas I don't know about, but not in the parts of industry I work in.

I would be curious to know about any research demonstrating significant industry-wide software quality improvements. As I understand, it's been pretty stable in terms of bugs-per-SLOC over the decades, but my knowledge is dated.

Reply on news.ycombinator.com