Conversation

Notices

1. Christine Lemmer-Webber (cwebber@octodon.social)'s status on Thursday, 29-Dec-2022 03:21:42 JST Christine Lemmer-Webber

   Twice a month someone shows up in the #spritely channel saying "I'm going to rewrite Goblins in Rust!", usually without having used it yet (also not realizing that a couple of common Goblins patterns don't have a clean mapping onto Rust at present)

   I love the enthusiasm of Rust's userbase, but...

   • Haelwenn /элвэн/ :triskell: (lanodan@queer.hacktivis.me)'s status on Thursday, 29-Dec-2022 03:21:38 JST Haelwenn /элвэн/ :triskell:

     in reply to
     @cwebber Turing complete isn't really enough, it also needs to be general purpose, be friendly for what you're doing and have existing tools/libraries to help.
     Unix shell is my preferred example of an incapable language for many things as it's pretty much a domain specific language.
     But C, Python, Elixir, Go, … also are depending on what you're doing.
     
     As for Rust, yeah, I wouldn't be surprised that it would lack tools and libraries for what Goblins is trying to do.
     For example the lack of a rust-native cryptographic library is probably not going to help (rust stdlib just binds to OpenSSL).
   • Christine Lemmer-Webber (cwebber@octodon.social)'s status on Thursday, 29-Dec-2022 03:21:40 JST Christine Lemmer-Webber

     in reply to

     I don't think I expressed what I wanted to well here, based on the replies. This was not meant as a dig at Rust or to declare it incapable. I think it's done awesome things. It was really a combo of the following:
     - Admiration for the enthusiasm of people in the Rust community to *implement*
     - A wish that people would try Goblins first before declaring an intent to reimplement
     - An expression that, while all paradigms are implementable from all turing complete languages, certain languages are tuned to solve certain problems in certain ways and that affects what feels *comfortable* to do.

     Adrian Cochrane repeated this.
   • Christine Lemmer-Webber (cwebber@octodon.social)'s status on Thursday, 29-Dec-2022 03:21:41 JST Christine Lemmer-Webber

     in reply to

     And by that I mean a couple of really core patterns. I usually say "try to write a general-purpose revokeable proxy *without* using code generation, that can work with any types" as an exercise for someone to get it.

     (It's not that it would be impossible to build a type system which could do static analysis and support that, but Rust's doesn't. I have a sketch for a propagators-based type system that might shatter the false dichotomy between static and dynamic strongly typed systems but that's a far future thing and not my current fight.)