miniblog.

RISC-V state of the union: Linux, binutils, GCC, LLVM, CompCert and OCaml support! Government funded projects in India, Israel and USA! https://www.lowrisc.org/blog/2017/11/seventh-risc-v-workshop-day-one/ It's impressive how much momentum the project is developing.

It's fascinating to learn that we're still finding new ways of making primality tests faster.

CLIs are reified UIs: https://www.expressionsofchange.org/reification-of-interaction/ Interesting exploration of what properties a text interface provides, and what it means for computer interaction.

Easier to read than write! I can't comment on Wolfram (I haven't used it), but this is a great goal to strive towards when coding. I've definitely experienced programs and languages where it's easier to write than read though.

Really excited to see Remacs exploring bindgen as an option! https://github.com/Wilfred/remacs/pull/486/ The C parts of Emacs use a ton of C features, so it's shaking out several bugs in bindgen itself.

A big milestone for Remacs: we now have CI running on Windows too! https://github.com/Wilfred/remacs/pull/445 I can't take any credit for this: Jean Dudey has done an incredible job.

How Elm slays a UI antipattern: https://blog.jenkster.com/2016/06/how-elm-slays-a-ui-antipattern.html Argues for GADTs in web dev, and points out a common issue with async UIs that I can't unsee now.

It's amazing how important kinaesthetic learning is in programming. There have been many concepts that I haven't fully grokked until I've actually coded it up myself. It forces you to consider all the details.

It's easy to observe software getting slower over time. It's nice to periodically reflect on software features that have become faster: sound codecs, JS execution, CSS engines.

Python is the only language I've seen where the stdlib docs contain smilies: https://docs.python.org/3.6/library/heapq.html#theory It makes the docs more approachable, but also shows the diversity of writing styles used!

Shower thought: suppose you have two groups of devs, one more experienced in a PL than the other. Ask everyone to solve a programming problem. I suspect the more experienced group would have more similar solutions. What things would they do similarly, and how would they differ?

Porting 600KLOC of Closure Compiler JS to typescript: https://www.lucidchart.com/techblog/2017/11/16/converting-600k-lines-to-typescript-in-72-hours/ Although Closure offers a set of static analysis tools, typescript is often stricter! This finds bugs but complicates porting.

Hyper-G is (was?) an interesting hypertext design, where hyperlinks were stored in a separate database to ensure integrity and allow visualisations with trivial backlink calculations: https://www.ickn.org/elements/hyper/cyb20.htm

Racket has a notion of a 'box', which is a single-element vector for when you want a mutable store of something: https://docs.racket-lang.org/reference/boxes.html Interesting, I've rarely seen this in other language as a first class concept (I can only think of ML).

Microsoft patching a compiled binary to fix a security issue: https://0patch.blogspot.co.uk/2017/11/did-microsoft-just-manually-patch-their.html Adding an additional parameter without changing the instruction count of functions is no mean feat!