1: Yes, I think paths are a big deal. path.join is more or less good enough. Everyone going around convincing each other that it's not
-
-
2: actually very important ("windows supports `/` yo" "uhhh not with verbatim paths") has been a huge huge regression.
4 replies 0 retweets 1 like -
3: But I think you're underestimating epoll/kqueue shimming, compiler/linker shimming, OpenSSL/crypto shimming, etc.
2 replies 0 retweets 1 like -
4: OSX and Linux share a path model (but not exactly a file system model), but they diverge a huge amount in general.
2 replies 0 retweets 0 likes -
5: the reason most people don't perceive that is that a lot of "Unix" libraries are actually just "shimming the union of OSX and Linux"
1 reply 0 retweets 2 likes -
6: epoll and kqueue are a good enough exemplar of this phenomenon to drive a discussion about this, but there's way more.
2 replies 0 retweets 0 likes -
7: for a fun time, check out the practical divergences in the Rust mio library (equiv of libuv): https://github.com/carllerche/mio/blob/91f0832036303c7f83c286e049e31be29c43b519/src/sys/unix/mod.rs …
1 reply 0 retweets 0 likes -
8/8: In practice, getting enough core infra to support all of Linux/OSX/Windows is harder than just Linux/OSX but important.
1 reply 0 retweets 0 likes -
Replying to @wycats
It’s not just harder. It’s impossible to do without leaking to the app level significantly, if the app touches paths or signals.
1 reply 0 retweets 0 likes -
Replying to @izs
Paths can be largely abstracted, signals shouldn't be used directly, but something approximating the signal mental model is fine.
1 reply 0 retweets 0 likes
"trap SIGINT" is worse than "handle control-c" but shimming the Unix mental model into Windows (for the common use-cases) is probably fine
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.