Convergence options for Ecmascript/Actionscript?
brendan at mozilla.org
Wed Aug 29 08:55:54 PDT 2012
Totally cool to push static, not a bone of contention. What you write
here and in reply to Alex for me cements the point that we should not
treat "AS3" developers as potential hand-coders of ES6. Alex is right,
Flash developers are not ready to switch programming languages *and*
rendering, styling, etc. APIs and other languages (CSS) all at once.
What's more, for AS4 to ES6 to hit the performance optimized spots does
not require ES6 changes so much as measuring on top implementations and
looking in white boxes (open source impls).
Feedback on implementors might help, but aligning with their main
concern, making the JS patterns actually used on the web (and by stupid
old benchmarks, too :-/), is the best way to get high perf AS4-to-ES6
Avik Chaudhuri wrote:
> As one of the designers of AS4 (the next version of ActionScript), I can say that we are looking mostly to (1) remove awkward ES4 features that come in the way of simplicity and performance, as well as (2) limit dynamic features of the language and boost static features at the same time. (We have our reasons, and I don't want to go into them here.)
> Of course, a part of Adobe is also devoted to HTML5/JS efforts, but at this point they are disconnected from ActionScript evolution.
> On Aug 28, 2012, at 5:30 PM, Brendan Eich wrote:
>> Claus Reinke wrote:
>>> Hope this isn't politically inappropriate here;-)
>>> 1. Flash is dying as a browser plugin, but otherwise still alive,
>>> especially with compilation of Actionscript to native code [2,3].
>>> 3. Actionscript is based on an old ES draft standard that failed to
>>> reach consensus, partly because its changes were considered
>>> too much at once. [my impression - I don't have a reference]
>> You really should read back in es-discuss if you have time (understand
>> if you don't!). We covered what made ES4 fail. The main problem was
>> namespaces, upon which packages were built.
>> Unfortunately, AS3 uses namespaces and packages heavily. Mozilla's
>> Shumway project includes an AS3 bytecode recompiler that generates JS,
>> and we cannot lower namespaces to anything native and JIT-optimized in
>> JS itself. Cc'ing Tobias in case he can comment.
>>> 4. Lots of Actionscript developers find themselves having to
>> Developers do adapt, of course.
>>> X, for Flash developers". When I read such blog posts (eg, , with
>>> side-by-side comparisons of AS and JS code), I get the feeling that
>>> the transition would be much easier from AS3 to ES6, where things
>>> like modules and minimal classes exist. There is an opportunity here,
>>> if those parts of the standard are sufficiently stable to recommend
>>> transpiling (until real ES6 implementations come along).
>> Namespaces and packages are the problems (really one problem, namespaces
>> making lookup three dimensional rather than two dimensions (proto and
>> scope chain).
>>> One consequence of 3 is that there should be a vibrant community
>>> of AS developers who have practical experience with features similar
>> Not similar enough, in my view. Not only does AS3 have namespaces, it
>> uses them for class-private and IIRC protected visibility. IIRC AS3
>> classes do not have class-side inheritance (just like Java), another
>> Finally, AS3 was really intended, along with an implicitly early-bound
>> namespace, to statically judge the meaning of names and even types. Its
>> code loading and global object model is entirely unlike the Web's.
>>> Yet I
>>> do not see any experience reports or spec feedback based on that
>>> here on this list. Wouldn't it be very helpful to invite such input?
>> I think not. Some of us were involved intimately in ES4. We have the
>> scars and remember the differences. ES6 does not need more constraints
>> or feature demands from newcomers. That is a sure way to kill it.
>>> I don't expect Actionscript developers want to lose any of their
>>> language features, but there might be profit (for both AS and ES
>>> developers) if Adobe could try to align AS4 to ES6 so that the
>>> common language subsets have common syntax and semantics.
>> That's up to Adobe, but I recall that they are talking about going more
>> their own way:
>>> Such a hypothetical AS4 would be a typed version of ES6, making
>>> it easier for Adobe's tools to support both, and making it easier
>>> for developers to use or maintain either.
>> No typed (static) version can work interoperably and soundly with the
>> dynamic language. This was one of the conclusions of the ES4 work. How
>> to interface typed and untyped (dynamically typed) code is still
>> research. Sam Tobin-Hochstadt's Typed Racket work is one example of
>> research along the lines of interconnecting dynamic and static code
>> without losing soundness.
>> Even throwing out soundness may break interop. We'll see how this works
>> for Dart, but right now Dart is a dynamic language with an early warning
>>> Are there any Adobe folks listening here? Does anyone have the
>>> contacts to raise the question with them? Do you agree that there
>>> is potential for useful exchange of information and joining of
>>> efforts if the wall between AS and ES could be lowered?
>> Adobe folks departed the Ecma TC39 field, although IIRC Adobe remainds a
>> member of www.ecma-international.org.
>> Again I must comment that ES6 wants no new constraints and pressures.
>> Scope creep is the enemy.
>> Also, you should not ignore the ES4 lessons learned or assume the
>> problems were not material and inherent to the divergence between AS3
>> and ES3.
>> es-discuss mailing list
>> es-discuss at mozilla.org
More information about the es-discuss