js0

[bkil]

Would you accept contributions that would show conformance to a much smaller subset as well (i.e., the smallest common denominator) and for refactoring some of the existing tests in making less assumptions about the runtime (to reduce false negatives)?

[ivankra]

That sounds fine to me, I'd happy to merge it.

Early JavaScript versions were basically defined by Mocha/SpiderMonkey features, JS1.0 = original Mocha (Netscape 2.0), JS1.1 = Mocha in NN 3, JS1.2 = SpiderMonkey 1.2, etc

Not sure about js0 - it could get subjective what is the common denominator. But js1.1 test suite that showcases exactly what Mocha could do (minus stuff/bugs that got dropped later) should be a fun one, especially since there's a binary to test against. Probably can't really do js1.0 - there's no source for old Mocha / Netscape 2.0 on the web, although there are binary releases.

[bkil]

That sounds like a promising opportunity for cooperation. So, you may have heard about the work in progress research I'm doing in this area (gemiweb0) and indeed the current subset of js0 is not final yet, but there are certain traits that feel pretty solid so far. I'm constraining myself to a tiny subset of ES3 which can be interpreted within a few thousand lines of code in the working PoC (or under 1k if implemented in a high level language).

My master plan is to focus on all FOSS js engine that is still maintained, especially those containing very few lines of code and to simultaneously patch out any false negatives from the test sets while also contributing small fixes upstream for any possible omission or error. I believe we could learn from such patching rounds which features and syntax is the least expensive so we could finalize the language based on that.

Then finally, we could submit the small set of lightest and best audited js engines as a component to every maintained tiny web browser that exists with a compilation (or runtime) flag to choose between engine.