Jonathan Blow

you pick awk as one of the few things you rewrite ... it doesn't make sense to me, because the real problem is the unix philosophy, not awk. 'st' seems alright, except the fact that it's limited to X-Windows is again a big problem (and probably drives complexity).

Looking more, I see that sbase has a lot of programs in it (it was not obvious from the page), but ... again it's just replicating the unix complexity but making it simpler. That's fine, but the real problem is the unix complexity.

I'm not really sure if Unix APIs are the reason why they have so few projects. I always assumed it's more like typical side project dynamics where people come and go and at some point the whole party is over.

I don't just mean the APIs, I mean, the fact that there are 69105 different command-line programs that all do tiny jobs, that talk to each other through weakly-typed, error-prone text. Those should be API calls in a library, not executables.

I would say that most of the Unix tools make sense for *interactive* usage, where errors caused by text interface are more immediate and apparent. The things get hairy when people try to immortalize them in shell scripts without handling any edge cases.