MathmagicianI took a quick look at the first draft, you seem to have made
similar decisions to what I did for my variant UCI spec.

From the miscellaneous questions: how large the numeric arguments can be is
actually somewhat important. I had my maximum supported hash size as 2 ^ 32,
but that broke cutechess.

For multiple search limits, my engine stops once any of them are met.

En-EnInteresting. I dug up the original which has proved very useful for
designing my own interfacer. The FSM diagram is particularly invaluable. I'll
double-check it a bit later and see if there are any comments I feel the need
to make.

MathmagicianOne thing that would be good (if possible) would be a program to
test compliance with the standard.

toanthThat would be amazing. But probably a lot of work I guess.

MathmagicianYeah, and it might not be possible to test every part of it.

expoYes! this is my actual ultimate goal – the thing I actually care about
making. And I think it's doable. (Maybe not testing compliance with every part,
but with most parts.)

TwipplySF fails.

analog horsObviously impossible. SF passes by default because whatever spec
becomes popular must be built to make its behaviour count as correct.

Twipply"I am the standard" – SF.