IRC channel logs
2019-09-13.log
back to list of logs
<OriansJ>Hagfish: with the goal of letting others contribute in parallel to reduce the load on individual developers. (that way we don't burn out janneke or anyone else here) <Hagfish>yeah, actually managing the intensity is the hard part <Hagfish>in a "do-ocracy", it's hard to hold someone back if they have the enthusiasm and the skills, and get really attached to solving a certain problem <OriansJ>Hagfish: plus everything is now in a state where what was once laughed at as impossible is really close to done <Hagfish>are there questions about which path is the best to take, or just questions about how hard each step is? <OriansJ>Hagfish: well, in the end; we now know that writing a simple C compiler is easier than writing a proper lisp <OriansJ>So, the question is there a best path; I think is answered by saying writing a C compiler in assembler got to the goal faster than anything else <OriansJ>(Mind you this is coming from a guy who write a classic garbage collected Lisp in Assembly first) <OriansJ>If we had more assembly programmers and a better base lisp; the answer could easily been quite different <Hagfish>i guess people are welcome to try different designs later, if they want to scratch that itch, and after this first path has shown what's possible <Hagfish>hopefully showing the proof of possibility will attract more people than the idea itself did <OriansJ>cc_x86.s being 4395 lines of highly isolated assembly functions and lisp.s being 3414 lines of highly integrated assembly functions. <OriansJ>I hope someone someday comes along and does the entire bootstrap better than I could have ever dreamed <OriansJ>I just hoped to provie that it was possible and didn't require magic or anything complicated to achieve <OriansJ>M2-Planet was designed to be understood in less than 30minutes as it is a trivial state machine <Hagfish>wow, yeah, that's something that someone could think through in their lunch break <Hagfish>i do think it would be a pity if some later iteration of the bootstrapping project ended up taking all the credit away from the work you've all done already <Hagfish>like poor old Stallman, telling people to call it "GNU/Linux" <Hagfish>i expect people to write epic poems about you <OriansJ>I never wanted to be famous, only to make the world just a little bit better for everyone who cared about trust and freedom <Hagfish>yeah, fame is a bad motivation to have, but i want a society that recognises and celebrates the work of selfless people <OriansJ>or me assuming guile is able handle (define foo bar) (define bar 3); in that order <OriansJ>goal for tomorrow morning after I wake up: get hex2 in scheme working-ish <janneke>the current toolset has enabled us to create a bootstrap path from mescc onwards <janneke>the M1 and hex2 formats are really helpful and performance of mescc-tools also helps a lot <janneke>i will be looking at removing all non-scheme tools frome the bootstrap path of mescc onwards except for mescc-tools <janneke>so your plan is a perfect match, i think <OriansJ>janneke: I think I can get 70-80% of the speed of mescc-tools in scheme <janneke>OriansJ: you can probably use read-char (and avoid textual-ports) <xentrac>OriansJ: your slow-utils plan sounds reasonable to me <OriansJ>janneke: good call on the read-char; one less dependency to deal with