[Win32utils-devel] RubyConf 2007 - Thoughts?

Daniel Berger djberg96 at gmail.com
Sun Nov 11 11:49:39 EST 2007


Hi Patrick!

Patrick Hurley wrote:
> On Nov 8, 2007 10:28 PM, Daniel Berger <djberg96 at gmail.com> wrote:
>> Hi,
>>
>> For those of you that attended RubyConf 2007, what were your impressions?
>>
>> Any more thoughts on JRuby, Rubinius, IronRuby or Wuby?
>>
>> Regards,
>>
>> Dan
>> _______________________________________________
>> win32utils-devel mailing list
>> win32utils-devel at rubyforge.org
>> http://rubyforge.org/mailman/listinfo/win32utils-devel
>>
> 
> JRuby is moving ahead with amazing speed. There current version
> appears to be on par or slightly faster than MRI, and their beta
> compiler will provide additional speed and the ability to lightly hide
> Ruby code for commercial deployment.

I have two problems with JRuby. First and foremost is the complete lack 
of an FFI. The whole "Just rewrite it in Java!" mentality that some 
people have is simply impractical. Second, the spec is nowhere near as 
complete as they would have you think.

Check out my personal test suite:

cvs -d :pserver:anonymous at rubyforge.org:/var/cvs/shards login
cvs -d :pserver:anonymous at rubyforge.org:/var/cvs/shards checkout ruby_test

Then try running JRuby against it. And then consider that I've 
explicitly commented out some tests for JRuby because it doesn't support 
certain methods. Like most of the Process module functions, for example.

> Rubinius is a sound approach and
> appears to be gaining a relatively large community, but it remains a
> work in progress -- I have high hopes, but there is a lot left to do.

I really, really, really wish Evan hadn't gone this route. Sydney was so 
promising, and actually had some cool new features. I don't see any 
practical advantage to targeting a two year old spec. Nor do I see 
Windows support materializing in the next year. Or two.

> IronRuby is still pretty early in its cycle as well, it appears to be
> the long term best choice for Windows and Ruby, but is at least a year
> from being usable (my wild guess not theirs) - it is encouraging that
> they have a reasonable license and are hosting on RubyForge, but it
> does not appear that they much community support yet (I guess
> microsoft has to work to gain open source developer trust, go figure).

Yeah, I actually considered applying for a position working with John at 
one point, but I just can't bring myself to work for Microsoft. I have 
considered applying for 501(3)(C) status for the Win32Utils project, and 
seeing if Microsoft would contribute.

> My overall impression is that for long running (non-script)
> applications JRuby is pretty much ready and developing rapidly. For
> everything else we still have one real Ruby. 1.9 is scheduled for
> Christmas this year, but Matz hedged his bets on its stability, but
> still encouraged everyone to use it, as without users it will never
> get stable.

I'm really not too happy with what I'm hearing about 1.9. So, we've 
added a VM, and a GIL. And a bunch of highly dubious methods to the core 
classes. I just don't see the ROI, especially when I think we simply 
could have made some really good improvements in the 1.8.6 branch. 
Anyone remember Eric Mahurin's array.c patch? I still want to get that 
into the 1.8.x branch.

http://blade.nagaokaut.ac.jp/cgi-bin/scat.rb/ruby/ruby-core/5861
http://blade.nagaokaut.ac.jp/cgi-bin/scat.rb/ruby/ruby-core/5869

And I've seen some interesting work on the garbage collector:

http://izumi.plan99.net/blog/index.php/2007/10/15/making-ruby%e2%80%99s-garbage-collector-copy-on-write-friendly-part-6-final/

> As for Wuby, as a packaging -- basically taking the one click
> installer, adding in some of the Win32 gems and possibly building with
> a recent version of VC (along with building all of the open source
> support libraries as well). That may have some value, but in the long
> term IronRuby is probably a better bet -- but that is at least a year
> off.

I'd like Wuby to be more than packaging, though. I'd like complete 
rewrites of some of the core classes.

Any, thanks for your input Patrick. Much appreciated.

Regards,

Dan


More information about the win32utils-devel mailing list