On 2009.05.13, at 11:38, Jason D. Clinton wrote:
On Wed, May 13, 2009 at 10:23 AM, Alberto Ruiz < aruiz gnome org> wrote:
That sounds to me more of a counter argument, is the GNOME official
desktop release the right place for a _javascript_ engine battle? Isn't
the performance of both already good enough for our purposes? It seems
to me that they both already perform better than Python and we've been
supporting it for ages already.
As has already been stated elsewhere, (with exception to the 'let' statement) they should be runtime compatible. The "battle" will happen regardless of what GNOME does.
Well, they're almost-runtime-compatible in one direction (GJS code running on Seed is more or less just missing 'let'), but I'm pretty sure there's a lot that has to be done in the other direction (I'm pretty sure that GObject subclassing (which we use a lot) isn't currently supported in GJS, among other things).
Performance-wise, it really appears that we're down to the "eke-ing out the last few percent" stage... I don't think there have been any significant performance gains since the V8 code dropped (and was promptly - at least partially - shoved sideways into JSC). In this case, it would help in the sense of having more "eyes"
focusing on one JS implementation--I agree. OTOH, if we are willing to
forgo that benefit in favour of hedging our bets in the battle for the
fastest, most featureful JS virtual machine, we gain a heterogeneous
JS platform that's robust enough to withstand the growing pains.
Featurewise, it's pretty clear that MozJS is going to be the leader for the foreseeable future, because the WebKit team is incredibly picky about performance, and are resistant to adding new features if they'll slow anything down. However, it's not clear (to me, at least) how many of these features are things that we actually care significantly about.
They are certainly mostly _not_ things that people coming from browser-based JS will know anything about or care to use (especially because they're already incredibly sensitive to incompatibilities), so the choice either way won't affect us on that particular front - those people are already accustomed to writing code that works on both MozJS and JSC. Note taht _javascript_ can be a potential entry vector for the GNOME
platform (which is one of the most interesting points of getting a JS
engine in), I think people will have a hard time to make a decision
they might not fully understand (the engince choice), and
documentation will get messier. Not to mention that we will end up
with one extra dependency.
Hopefully everyone can agree on some solution WRT to "let" and then we don't have to care which one they develop with: their work will run on both.
|
