Congrats Steve! Knock down those barriers to jj adoption one at a time!

P.s. building a code collaboration is good fun with interesting challenges, I’m sure you’ll enjoy it.

Can confirm that I’ve had nothing but problems with Spotify support. Dont worry though it’s not a sign of beef or anything, @danimalcannon.com and I are cool 😎

lol, and when I implemented my own public wrapper over it, and implemented `publicJWKFromPrivateJWK` the AI auto-complete confidently returned the private JWK as-is, asserting that the library "takes care of it for you" which AFAICT is NOT the case... the next generation of app security is cooked.

Even the #Haskell `jose` library doesn't have a separate type for public vs private JWKs, even a phantom type would do the trick.

TBH this is just as bad as how generated certs are named key and key.pub;

If you tab-complete you'll get the private one every time and there's no `.private` as a hint

At the moment I use STM as my lowest-level primitive; then leave it up to the STM implementors to use the mutexes haha.

STM has all the channels, queues and vars you need to do any form of actors or CSP you like :)

My current thoughts: go down the list and use the first approach that you can fit your problem into:

* Pipeline processing
* CSP
* Actors (a half dozen or fewer, statically defined)
* STM
* Mutexes (only ever use these if you're implementing STM 😆)

P.s. try clicking the HAL9000 in the header;

Error-as-value;

Yes, all tasks should be cancelled, `ki` takes care of that in my case.

hackage.haskell.org/package/ki-1...

hackage.haskell.org/package/ki-1...

Stay tuned and you won't be disappointed, Arrows and Category classes are next!

Oh whoops, lol that’s an important “not”!