intertwingly🔒 SSL Verified

Table of Contents

• ↓ Nav Bar
• ↓ Top Alternatives to intertwingly.net

6-Month Traffic Trend Estimate

Official Contact & Profiles

Visual representation extracted from intertwingly.net

Content Snippets from intertwingly.net

"Two days ago I wrote that the Roda spike's verdict wasn't mine to deliver — the results sat in front of Jeremy Evans, the keeper of this oracle. The verdict came back Tuesday evening: he built and ran both generated applications himself, independently reproduced the memory numbers, filed the kind of review that finds what CI can't — and then answered a question I'd posed about reversing the arrow. A translator from Rails to Roda/Sequel, he said, would be "both interesting and useful" — and in the same comment he set the usefulness bar, the failure policy, and the acceptance test. Wednesday morning, between his comment and lunch, the converter existed: --target roda turns the Rails blog into a Roda + Sequel application that runs on his real gems, his round-trip test passes as a committed gate, and his exemplar's own nineteen-check suite scores 17 of 19 against the output — with both failures turning out to be places where two hand-written, supposedly domain-identical applications genuinely disagree. The intermediate representation that last week proved it wasn't secretly Rails-shaped is now load-bearing in both directions: not just compile Roda too, but translate between frameworks, deterministically, with an equivalence test."

"In June, Matz spent a week rewriting Spinel — his ahead-of-time Ruby compiler — from self-hosted Ruby into C, because whole-program type analysis over the compiler's own 80,000 lines had pushed builds to twenty minutes. In July, the same analysis met a 40,000-line application and took two hundred seconds — because the rewrite had faithfully translated its O(N²) into a faster language. Over two days and five upstream pull requests, that two hundred seconds became five — including the discovery that the inference fixpoint had never once converged on a non-trivial program; every compile ran to its 128-iteration cap, and the cap made it terminate anyway. Every fix was the same bug, one loop deeper. The moral is not that the rewrite was wrong: the rewrite bought the fast native dev loop and clean profiles the algorithmic work was done on. You can't rewrite your way out of big-O — but the rewrite can build the workbench you'll need when big-O comes due."

"After my RubyConf talk, Jeremy Evans asked me a question: could you do Roda too? Four days and a few elapsed hours of work later, an idiomatic Roda + Sequel application — an exemplar Jeremy reviewed line by line before any compiler code existed — transpiles to the same intermediate representation as its Rails twin (93 of 130 lowered methods byte-identical), passes its own test suite unchanged, and compiles to a 559 KB native binary — via Spinel, Matz's ahead-of-time Ruby compiler — that boots to its first served request in ten milliseconds at 4.5 MB of memory. At request time there is no Roda, no Sequel, no ORM, no router — every query left the building at build time as a SQL string. The frameworks melted away, which is what frameworks are supposed to do when you point a specializer at them. But the part I most want on the record is the seating chart: I recently wrote that maintaining the oracle is the human job — and in this episode the keeper's seat isn't mine. Jeremy extended the spec where it was blind, decided how correctness gets computed, and holds the verdict on whether this spike lives, dies, or gets tweaked. The deepest open problem in that earlier post was behavior for which no reference exists; "idiomatic Roda" is exactly such a behavior, and the solution turned out to be social — the reference is a person."

Key Topics & Focus Areas for intertwingly.net:

Nav Bar