LightNews
rntz
@rntz.net
330 followers 170 following 280 posts
Michael Arntzenius irl. Postdoc at UC Berkeley doing PL + DB + incremental computation. PL design, math, calligraphy, idle musings, &c. rntz.net 🐘 @[email protected] 🐦 @arntzenius Attempting to use bsky more now that people are showing up.
Posts Media Videos Starter Packs
rntz @rntz.net · 1d
exactly! the table stores only the inputs mapping to non-nil.
1
rntz @rntz.net · 1d
without the nil value I can't define what finite support is. or am I misunderstanding the question?

(the slides are probably not self-explanatory without the talk, but I thought I'd throw them up anyway...)
1 1
rntz @rntz.net · 1d
bsky.app/profile/did:...
rntz @rntz.net · 1d
Slides for my HOPE 2025 presentation, "Finite Functional Programming via Graded Effects & Relevance Types": www.rntz.net/files/hope-2...
a handwritten slide, saying: lambda: the ULTIMATE RELATION! -> RELATIONS = FUNCTIONS + SUPPORT -> FINITE support = DATALOG. ENUMERABLE " = PROLOG ?? -> naturally handles: { WEIGHTED logic programming { AGGREGATIONS -> type system is UGLY but I love it <3 GRADED (CO)MONAD + ADJUNCTION + RELEVANT TYPES + HAX
1
rntz @rntz.net · 1d
Slides for my HOPE 2025 presentation, "Finite Functional Programming via Graded Effects & Relevance Types": www.rntz.net/files/hope-2...
a handwritten slide, saying: lambda: the ULTIMATE RELATION! -> RELATIONS = FUNCTIONS + SUPPORT -> FINITE support = DATALOG. ENUMERABLE " = PROLOG ?? -> naturally handles: { WEIGHTED logic programming { AGGREGATIONS -> type system is UGLY but I love it <3 GRADED (CO)MONAD + ADJUNCTION + RELEVANT TYPES + HAX
2 1 7
rntz @rntz.net · 2d
why am I calling my secret project "Finite Functional Programming"
when I could be calling it "Lambda: the Ultimate Relation"
1 7
rntz @rntz.net · 2d
no, we're all at the workshops at NUS School of Computing

join ussssss
rntz @rntz.net · 3d
why is writing (anything) so much like squeezing water from a stone?

approximately the most important thing in my profession. I have done it repeatedly. I wrote a whole thesis. it is no easier now than ever. perhaps harder. ughhhh
8
rntz @rntz.net · 3d
i go halfway around the world for a conference in a famous tourist city and first day what do I do?

go to a bookstore
two books lying on top of one another: Driftglass by Samuel R. Delany The Murderbot Diaries, Vol. 1 by Martha Wells
1
Reposted by rntz
joshuahhh.com
Josh Horowitz @joshuahhh.com · 8d
Does anyone here know a lot about monads & applicatives & such?

I'm curious about how to make spreadsheet-style data-flows "monadic".

If that sounds interesting please look at my scattered notes at typst.app/project/rnf1... and lmk what's going on.

Quick motivation in thread...
3 1 8
rntz @rntz.net · 9d
Cryptoconchoidsyphonostomata or WHILE IT'S TO BE HAD.
rntz @rntz.net · 12d
I wasn't thinking of synthetic functions, but of synthetic values. Eg. the delta(0) = infty, delta(x) = 0 otherwise, where "infty" is a synthetic value such that integrating over it yields 1. So infty is like the opposite of an infinitesimal. Indeed, perhaps infty = 1 / epsilon?
1
rntz @rntz.net · 13d
Analytically, we differentiate via limits/epsilon-delta. Synthetically, we use infinitesimals - artificial infinitely small quantities.

Analytically, we handle dirac deltas with distribution theory. What's the synthetic account? Is it artificial infinitely big quantities?
1
rntz @rntz.net · 16d
guess the subject of the paper
2 2
rntz @rntz.net · 17d
productivity of co(inductive)-programs
fairness of concurrent scheduling
completeness of search strategies

these three seem deeply related to me, but I can't yet precisely articulate how. Is there existing work on connections between them?
1 3
rntz @rntz.net · 18d
just a bit onward Hansen cites Dijkstra 1971b, "Hierarchical Ordering of Sequential Processes", which invokes a fairness property ("each process... is guaranteed to proceed with some unknown, but finite speed") but reserves the actual term "fair" only for a much stronger property
1
rntz @rntz.net · 18d
nice, that is earlier than anything I've seen. still, I bet it was just folklore / "in the air" rather than clearly originating in any one publication.
1 1
rntz @rntz.net · 18d
haven't found where it started but I did find a 2012 paper trying to precisely define the "class" of fairness properties, currently reading it: dl.acm.org/doi/pdf/10.1...

also Lamport took a stab at it in 2000: lamport.azurewebsites.net/pubs/lamport...
Defining Fairness in Reactive and Concurrent Systems | Journal of the ACM
We define when a linear-time temporal property is a fairness property with respect to a given system. This captures the essence shared by most fairness assumptions that are used in the specification and verification of reactive and concurrent systems, ...
dl.acm.org
1
rntz @rntz.net · 18d
where did the term "fairness" in concurrent programming originate? is there a canonical definition? a canonical citation?

I've found references in Owicki & Lamport 1982 & Andrews & Schneider 1983, but neither originated the term or concept.
2 1 1
rntz @rntz.net · 18d
what's the equivalent of nicotine patches, but for social media addiction?
2
rntz @rntz.net · 19d
imagine an inverse h-index (a ɥ-index?): not how widely an author is cited, but how widely an author cites. do they cite from other subfields? other fields? other centuries? other languages? etc.

who do you know who'd have a high ɥ-index?
1 7
rntz @rntz.net · 19d
typing out Haskell code from a paper by hand while listening to Trail of Dead like it's 2005
7
rntz @rntz.net · 25d
I've materialized the transitive closure of Slashdot Zoo social net circa 2008. Took 2m25s on a 128-core 500GB machine. Peaked at over 200GB mem usage and 6400% cpu util. The parts I haven't yet parallelized (LSM merging, concatenating partitions) took ~half the runtime. Amdahl's law strikes again!
3
rntz @rntz.net · 25d
two wolf inside me

one howl: SATURATE AVAILABLE PARALLISM! NEVER LEAVE A CORE IDLE!

other one growl: AVOID UNNECESSARY WORK. MINIMIZE COMMUNICATION OVERHEAD.
2
rntz @rntz.net · 28d
I got irritated at how tightly spaced EB Garamond 08 (the version intended for use at smaller sizes) is, so I tried my hand at re-spacing the 26 lower-case letters manually in FontForge. Here's the result (original spacing first, then my adjustments).

github.com/georgd/EB-Ga...
rntz @rntz.net · 29d
ad hominem / rad hominin