easy to fix

• heap/stack init is not clear
  mp_js_init(stack_size) with stack= parameter instead of heapsize=
  https://github.com/micropython/micropython/blob/master/ports/javascript/wrapper.js#L59
  but in fact is heap https://github.com/micropython/micropython/blob/master/ports/javascript/main.c#L88

• add more .exp test files so one can write a browser testsuite or use run-tests-exp.sh.

• browser example is minimalist and does not provide a default micropython web server and possibly a websockify example

• builtins time.sleep and input should at least throw a warning/error instead of crashing browsers.

• an asyncio flavour should be provided as a frozen module.

• please use cpython C-API names when interfacing JS and C that will save everybody's time. either for understanding api or writing documentation see https://docs.python.org/3/c-api/veryhigh.html. that would ease transition later to https://github.com/pyhandle/hpy

• provides simple sync I/O for snippets run cases POC port-javascript: add sync file access for scripts compatibility lvgl/lv_micropython#6

annoying things :

• not so easy to use a pure async stdlib out of the box because of basic discrepancies
  butilin iterator tool RFC: MICROPY_PY_BUILTINS_NEXT2 always set to (0) ? #5046

• not so easy to interface with browser nor subprocess ( proxies )

  class.mro() is notimpl RFC: class.mro() is notimpl #5106

• why isn't javascript port yet the experimental platform of choice avaiable to everyone ?
  nodejs has a sandbox mode equivalent to browser.
  and import of online module can be done.
  cf triaging : separate ideas-features-request from BUGS #4444

• passing script data for interactive or script should be separate functions, actually all use mp_js_do_str(text) and alloc context of text is not clear if pausing/resuming is needed. But those are not the same :
  https://docs.python.org/3/c-api/veryhigh.html#c.PyRun_SimpleString
  https://docs.python.org/3/c-api/veryhigh.html#c.PyRun_InteractiveLoop

• parser/lexer should work on a pre allocated buffer either for interactive or script mode, filled with stringToUTF8( text, pointer, max_size ) as a real C string.

  • why pre alloc ? because max size of code that interpreter can compile is fixed by heap limit anyway so save useless/unclear malloc/free operations beetween js/C
  • execution mode could be simply set at end of buffer in a # comment when using asynchronous execution.

• mp_js_do_str(text) is not wasm compliant and won't work when using clang 10 from latest-upstream emscripten branch. so better implement the two previous point and just get rid of it.

• javascript port can't blink a led.
  well not yet but soon https://wicg.github.io/webusb ;)
  apart from the usb joke it can via websocket and a networked board but there's no sample.

heavy lifting :

• not so easy to interface with browser nor subprocess
  non GUI threads are available on wasm but GC is not under stdlib control turning everything simple into a leak source :
  py: __del__ special method not implemented for user-defined classes #1878
  Do we want to have weakref support? #646

• internationalisation is deeply missing.

• extra features available as diff only.
  f-string py: Implement partial PEP-498 (f-string) support #4998
  assignment expressions py/compile: Implement PEP 572, assignment expressions. #4908
  => problem is that port is not meant to be user built and parametric, it is expected to be full options and on a CDN.

• files fd, sockets fd, pipes and not only stdin/stdout should be multiplexed via stdin/stdout via a common interface for node / browser and worker. zero copy rpc would be better ( arrow / capnproto ? )

  • why ? because MicroPython has ability to get a lot better than brython and cpython on the web.
  • do not expect WASI to solve any async problem specific to python main thread in the browser.
  • even if WASI get async, it's polyfill is not yet ready to load emscripten .wasm.

probably too heavy without branching/forking/redesign.

• micropython core should could be switched to non nlr with [proof of concept] completely remove nlr from py, extmod and unix port #4131. emscripten-upstream gets seriously confused with setjmp but fastcomp branch is to be dropped ( and anyway is troublesome with function pointers ).

UPDATE 05/09 : a fix is in progress https://reviews.llvm.org/D66784

• performance and I/O : asyncify is 10x slower than real thing : the solution is C-stackless design
  that's both solve asynchronous I/O on any port and Allow pausing and resuming execution #3619 ( also called "checkpoint and restore in userspace" ).