What was the motive in choosing algorithm description approach? about infra

[annevk]

I think in part because English prose is always the way it's been done, so formalizing around that seems less disruptive than inventing a syntax. But English prose also tends to be more readable and leaves more room for being creative in case you need to do something novel. Furthermore, if we had a syntax such as used in that paper, we'd then have to standardize that syntax fully and be constrained by it. Typically what I find is that such pseudocode is not actually defined in detail and therefore more ambiguous than English.

The reason why we don't go for performance is because that is really complicated. E.g., consider the JavaScript standard. If they had to concern themselves with performance they'd have to define JITs, a dozen string types, etc. For CSS you might end up having to define multiple selector matching engines and fiddle with them over time as the characteristics of the web change. It's much easier to describe input -> output and leave the bits between to be optimized as implementations see fit. That also allows completely novel implementation approaches such as Servo.

[domenic]

@annevk, I think the last paragraph there could be morphed into a useful non-normative note in the spec. Less sure about the first one.

[annevk]

@LeonidVasilyev are you okay with closing this? The standard now includes a note on performance. I'm not quite sure what to say about pseudocode other than using it meaning we'd have to define how that works (and how you extend it), etc.

[LeonidVasilyev]

@annevk yes, thank you.