MoarVM: A virtual machine for NQP and Rakudo | Hacker News Reader

MoarVM: A virtual machine for NQP and Rakudo

6guts.wordpress.com

40 pointskbenson23 comments

I just wrote a post about MoarVM, too: http://strangelyconsistent.org/blog/a-vm-for-6model . Slightly more slanted towards how this fits into the evolution of Perl 6 implementations.

Regarding "What achievement can Perl 6 claim?"

At least one is 6Model itself -- a generic object semantics and API which has been back ported to Perl 5, has been ported to the JVM (with a lot of optimization still pending) and as we see here is a reasonable design for part of a VM API.

As far as I know 6Model is the only candidate around for a low level object model that is flexible enough to implement the object semantics in most languages, low level enough to be be efficiently implemented in most VMs, etc.

If anyone is curious look at https://github.com/jnthn/6model/blob/master/overview.pod

There may be more comprehensive / current documentation out there but I don't know where.

a generic object semantics and API which has been back ported to Perl 5

Do you mean Moose or the abandoned Perl 5 MOP experiment?

Both, I guess, though claiming Moose is better PR.

I don't know enough history here. The current Moose docs say "Moose is built on top of Class::MOP" so maybe there was a different MOP effort that succeeded? That MOP is probably closer to 6Model than Moose itself.

Moose and Class::MOP predate 6Model. As far as I remember and can verify, Moose 1.0 was out (spring 2010) before 6Model even existed or had a name (late 2010).

Moose did take inspiration from the object model designed for Perl 6, but it grew out of attempts to implement that object model in Pugs, not Rakudo. You might even say that it was the first full implementation of the Perl 6 object model.

Thanks for the history. Sounds right. So 6Model and Moose are siblings, rather than parent / child.

I think my original point, though, holds up: One of the actual delivered achievements of the Perl 6 project is a sophisticated flexible object model, now in production outside of Perl 6, and also the basis of deeper (re)design within Perl 6.

Wow, they are really good at starting things.

Right... start things... keep evolving them and getting better... and hopefully never end.

I think that's the plan.

Never end? More like never deliver.

It seems to me they deliver a release every month.

http://rakudo.org/downloads/star/

Same as firefox... or chrome... or any other project on a continuous release cycle.

Firefox and Chrome have actual users who get stuff done, probably because their developers have a track record of releasing useful software. Perl 6 has a track record of not being useful, but I'm sure /this/ time they'll buck that trend.

Your belief that Perl6 won't go mainstream doesn't lessen their achievements.

Beyond that, your attitude is easy to compare to a 2000s "SSJS will never succeed".

If 13 years of futility still leave you optimistic, your mind is far more open than mine.

What exactly is the achievement here?

I'm not certain that yet another mostly-incomplete virtual machine will really help the Perl 6 community much. They've already got that with Parrot.

The whole emphasis on targeting virtual machines, whether it's Parrot, or the JVM, or now this MoarVM, has harmed the ability of Perl 6 to be implemented. We've seen one partial implementation after another, for years on end.

Unfortunately, we can't actually use any of these implementations for anything serious, like we can with Perl 5, Python, and Ruby. This makes Perl 6 unusable, which is quite a shame.

We've seen one partial implementation after another, for years on end.

That's the second most disappointing part. The Parrot developers had a strong push between two and three years ago to redesign the internals to be a better fit for Rakudo along these lines (better runcore, easier JITting, improvements to the core object model, representation polymorphism), but the reaction from the Rakudo developers was a mixture of "No, thank you!", "Not yet.", and "If you have to, but don't change any APIs or behavior that Rakudo counts on."

It's a shame this couldn't have all been done at the same time. The available volunteer knowledge and interest might have saved two or three years of Perl 6 floundering around in the netherworld of big promises and very modest realities.

The most disappointing part is realizing that this sets back the potential delivery date of a usable Perl 6 by at least another couple of years. I guess that shouldn't be a surprise by now though.

I watch the Perl 6 project closely. You are misunderstanding what is going on.

Perl 6 has to target a VM, just as Perl 5, Python and Ruby do. The Rakudo Perl 6 team picked the Parrot VM and tried to focus on everything but the VM. That's clearly no longer tenable because Parrot has turned out to be a Norwegian Blue (at least at the moment).

It's disappointing that Perl 6 still isn't ready for prime time. Fortunately Larry Wall and friends are changing that. I'm confident MoarVM is yet another big leap in the right direction.

Isn't that a different argument than the one you started making?

Brilliant! Let's /do/ have an argument about what "useful" or "production" mean.

If you're not even going to put the barest effort into keeping up the charade of anonymity, why bother posting from a fake account?

You may be trying to make a point with the account name, or may think you may not be able to get a fair conversation without the pseudo anonymity, but frankly, (I would hope) it's beneath you.

I would be happy to discuss my views on what useful and production mean. I'm pretty sure I know your current views on the topic (as a long time reader of your blog, my recollection is that they seem to have shifted over time).

Then again, maybe you aren't interested in having that discussion with me. If so, this seems a fairly inefficient way to communicate to the developers.

For what it's worth and as someone who took what turns out to have been the wrong side of those arguments, I apologize.

I was involved in Firefox marketing. If the actual Firefox story is the template, Perl 6 is poised for great success.

The project that led to Firefox was started in 1993. It was called Netscape. (Let's say this is analogous with Perl 5.)

In 1998 the project was reinvented as Mozilla. (Let's say this is analogous with Perl 6.)

For much of the 7 years between early 1998 and late 2004 the Mozilla project (analogous with Perl 6) was lambasted by the masses for, first, failing to release anything, and then, when they did, releasing slow, buggy, bloated, irrelevant software. By 2003 Mozilla was being ignored by almost everyone.

What saved Mozilla was that some smart engineers refused to bow to, on the one hand, public disinterest, and, on the other, internal pressure to focus on improving Mozilla 1.0. Instead they did yet another rewrite of several key parts, and then in early 2003 those still interested got behind the rewrite and the outcome was Firefox 1.0 in late 2004.

The parallels with the Rakudo team's efforts are clear and I see scope for an analogous outcome.

Reply on news.ycombinator.com