Twitter | Search | |
Gary Bernhardt
Illuminating the dark corners of programming. Destroy All Software (dense programming screencasts); Deconstruct (independent software development conference).
2,867
Tweets
109
Following
29,797
Followers
Tweets
Gary Bernhardt Jun 16
It is, yes. And the pipes aren't pure in our actual OSes, but we can imagine an OS and shell where they are.
Reply Retweet Like
Gary Bernhardt Jun 16
Replying to @arepty
Yep, 2007. It might've been 2006, but close enough.
Reply Retweet Like
Gary Bernhardt Jun 16
Replying to @searls
Nagle is completely the wrong thing here; I mean the congestion window. Forgive me; I'm at an airport bar with attendant drop in cognitive faculties.
Reply Retweet Like
Gary Bernhardt Jun 16
Replying to @searls
Was that drop consistent or on a cold request? Once a persistent HTTP connection is up and the nagle window has opened, I'd expect the two to perform similarly for request/response.
Reply Retweet Like
Gary Bernhardt Jun 16
Replying to @garybernhardt
Wait, I'm off by a decade. Smalltalk was 35 years old.
Reply Retweet Like
Gary Bernhardt Jun 16
Replying to @garybernhardt
A table full of dudes, all older than me (I think I was 24), all leaned in, genuinely flabbergasted that anyone could believe you can successfully build anything beyond simple scripts in any dynamic language. Smalltalk was already 25 years old then.
Reply Retweet Like
Gary Bernhardt Jun 16
Replying to @garybernhardt
This is a great time to reiterate one of my favorite stories. ~8 Microsoft-tech-using developers, plus me, at lunch at my first ever conference. Every one of them adamant that "real software" can't be built in Python, or any dynamic language. I wonder how many are JS devs now.
Reply Retweet Like
Gary Bernhardt Jun 16
Replying to @searls
Why was the xhr slow? Seems like Rails should be able to turn these simple requests around in a few ms.
Reply Retweet Like
Gary Bernhardt Jun 16
Replying to @samphippen
rubbyfbi
Reply Retweet Like
Gary Bernhardt Jun 16
Replying to @samphippen
you called
Reply Retweet Like
Gary Bernhardt Jun 16
A compiler doesn't need side effects either. `cat source.c | compiler`. No side effects!
Reply Retweet Like
Gary Bernhardt Jun 16
This is the greatest story ever written about "the programming opinions I've accumulated are Morally Good"... even though they're mostly the product of what you happened to be exposed to, which happened mostly by chance.
Reply Retweet Like
Gary Bernhardt Jun 16
The talk that Sam linked () is about exactly this type of belief, which falls apart almost immediately upon inspection.
Reply Retweet Like
Gary Bernhardt Jun 16
I'm not talking about asserting that something has the right class. I'm talking about, for example, testing for error handling in exceptional cases that a good type system can make impossible. Almost every test involving a null value is in that category.
Reply Retweet Like
Gary Bernhardt Jun 16
I don't think compilers are useless!
Reply Retweet Like
Gary Bernhardt Jun 16
Replying to @mgattozzi @ManishEarth
I should've clarified: what I want is to be able to build an app, compile to wasm for production, and never involve npm at all. No need to publish packages anywhere. Does that change it or am I still looking at npm in the short term with hopeful removal long-term?
Reply Retweet Like
Gary Bernhardt Jun 16
Replying to @ManishEarth @mgattozzi
What does the cultural metagame look like around that? Do you think no-npm will be tenable in a few years, or will that case end up mired in problems due to swimming against the current?
Reply Retweet Like
Gary Bernhardt Jun 16
Replying to @samphippen @sarahmei
This maps to Ideology as: you don't need to write tests for possibilities the type system excludes. The more powerful your type system, the bigger that space is. In Java you'll still test a lot; in Haskell you're not going to test much; and most are in between.
Reply Retweet Like
Gary Bernhardt Jun 16
What does the Rust-to-wasm situation smell like right now? Is the full path robust and well-used yet? Can it be done with no npm installed, while staying within a relatively well-used path?
Reply Retweet Like
Gary Bernhardt Jun 16
Replying to @bodil
You're doing Dijkstra's own work in this thread!
Reply Retweet Like