5 ms·

Rust Abstractions Posted for Sockets and Other Fundamental Network Bits

63 points by unripe_syntax 3y ago

cozzyd 3y ago
and the initial response seems to be, "why?" https://lore.kernel.org/lkml/0e91e3be-abbb-4bf7-be05-ba75c7522736@lunn.ch/ https://lore.kernel.org/lkml/0e91e3be-abbb-4bf7-be05-ba75c75...
- 38 3y ago
  the response Andrew Lunn seems reasonable. the Linux kernel is 30 years old. its pretty big already. do you have any idea the monster it would look like, if they just said "yes" to everything? I want people like Andrew and Greg looking critically at additions, not just rubber stamping everything. otherwise, the kernel will be reading email soon: https://wikipedia.org/wiki/Jamie_Zawinski#Zawinski's_Law https://wikipedia.org/wiki/Jamie_Zawinski#Zawinski's_Law
  - cozzyd 3y ago
    It's very reasonable indeed, considering the long-term maintenance burden of any API.
- chris_wot 3y ago
  It looks more like - we need something to be using it. Can you provide a program that uses it so we can evaluate this better.
- pjmlp 3y ago
  I am still surprised that Linus is perfectly fine with possible clever use of Rust macros, and derivation attributes, which are hardly different from his complaints towards C++ metaprograming.
  - zdimension 3y ago
    Having used both, I find it much easier to predict the behavior of Rust code than C++ code when metaprogramming is involved. Rust is very much against having thirty-seven and a half automatic deduction mechanisms under a trenchcoat and calling that metaprogramming. C++ has done a great job at building stuff on top of C but it's a bit like Windows 9x at this point, limited by DOS. C++ templates and resolution rules are fundamentally flawed from a UX perspective, however powerful they are. In the same vein, Rust usually has much better error locality, i.e. in C++ it's quite easy to get three pages of errors in files from the std when the error is actually in your code and you forgot a `const` or a `&&` somewhere. This pretty much never happens in Rust, or, well, other languages. This is HN so usual disclaimer, it's not always proselytism to find one language's way of doing things more well-thought than another when twenty years of programming language research separate them.
    - pjmlp 3y ago
      A rant that would be relevant, if concepts and constexpr did not exist.

63 points by unripe_syntax 3y ago

and the initial response seems to be, "why?" https://lore.kernel.org/lkml/0e91e3be-abbb-4bf7-be05-ba75c7522736@lunn.ch/ https://lore.kernel.org/lkml/0e91e3be-abbb-4bf7-be05-ba75c75...

38 3y ago
the response Andrew Lunn seems reasonable. the Linux kernel is 30 years old. its pretty big already. do you have any idea the monster it would look like, if they just said "yes" to everything? I want people like Andrew and Greg looking critically at additions, not just rubber stamping everything. otherwise, the kernel will be reading email soon: https://wikipedia.org/wiki/Jamie_Zawinski#Zawinski's_Law https://wikipedia.org/wiki/Jamie_Zawinski#Zawinski's_Law
- cozzyd 3y ago
  It's very reasonable indeed, considering the long-term maintenance burden of any API.
chris_wot 3y ago
It looks more like - we need something to be using it. Can you provide a program that uses it so we can evaluate this better.
pjmlp 3y ago
I am still surprised that Linus is perfectly fine with possible clever use of Rust macros, and derivation attributes, which are hardly different from his complaints towards C++ metaprograming.
- zdimension 3y ago
  Having used both, I find it much easier to predict the behavior of Rust code than C++ code when metaprogramming is involved. Rust is very much against having thirty-seven and a half automatic deduction mechanisms under a trenchcoat and calling that metaprogramming. C++ has done a great job at building stuff on top of C but it's a bit like Windows 9x at this point, limited by DOS. C++ templates and resolution rules are fundamentally flawed from a UX perspective, however powerful they are. In the same vein, Rust usually has much better error locality, i.e. in C++ it's quite easy to get three pages of errors in files from the std when the error is actually in your code and you forgot a `const` or a `&&` somewhere. This pretty much never happens in Rust, or, well, other languages. This is HN so usual disclaimer, it's not always proselytism to find one language's way of doing things more well-thought than another when twenty years of programming language research separate them.
  - pjmlp 3y ago
    A rant that would be relevant, if concepts and constexpr did not exist.