IRC channel logs

<homo>reading https://augustss.blogspot.com/2014/12/a-commentary-on-24-days-of-ghc.html https://blog.ocharles.org.uk/blog/pages/2014-12-01-24-days-of-ghc-extensions.html while it's questionable whether they belong in haskell or in separate language, it's strange to make bootstrapping harder by using them to implement compiler

<jackdk>If you care about bootstrapping, I think it strange to have self-hosted compilers at all. If you care about using more powerful languages to solve problems, depending on extensions is not really that different to depending on the language you're implementing, IMHO.

<homo>self-hosted with bootstrap path is not a problem at all

<jackdk>technically no, but practically it's tempting to lean into the language in ways that sever it from its bootstrap path - look at what happened to Zig.

<homo>never heard of zig other than its name

<jackdk>They had a compiler written in C++ maintained, and for a while it was maintained in parallel with the self-hosted version. They now compile the compiler to WASM, compress and commit that to the repo, and "bootstrap" the compiler by using the system C compiler to build a tiny WASM runtime that can run that Zig compiler blob. Very cool but very not a true bootstrap.

<homo>I'm surprised there are over 100 extensions, but I'm not using any of those when writing in haskell

<unmush>jackdk: sounds like mono.

<unmush>except that mono didn't even check their monolite blobs into version control to my knowledge, they just download them from mono-project.com at build-time

<unmush>anyway, as long as bootstrappable versions exist, and no version relies on features exclusive to unbootstrappable versions, the worst that can happen should be that the bootstrap chain gets really long

<homo>considering there are so many extensions, it seems like maintainers of other implementations got overwhelmed by the amount needed to implement, kinda like why there is no interest in having more independent web browsers

IRC channel logs

2024-12-10.log