From: Robert Feldt <feldt@...>
Date: 2001-11-28T06:49:31+09:00
Subject: [ruby-talk:26702] Re: Vote in the current poll!

On Wed, 28 Nov 2001, Phil Tomson wrote:

> >Basically do you think there is a market for a Ruby compiler and binary
> >tools? (sure it would be great if its free but if one wants
> >commercial-grade stuff I think someone needs to pay a little; maybe
> >only companies?) Or would you be content with a faster interpreter? 
> 
> Personally, I'd be content with a faster interpreter (VM?).  Compilers for 
> a language like Ruby seem to be somewhat problematic because of the 
> dynamic nature of the language - I'm not sure you would get as much of a 
> speedup as is implied.  
> 
IMHO, you can get a considerable speed-up. AOS/SmallScript is a case in
point (but I'm in the beta program and not allowed to give any
figures). Its not easy but it can be done.

> If your premise is that if we can show that people are building large 
> programs with Ruby  we can prove the need for [compilers|faster 
> interpreters] then I'm not sure that's the right way to go about it.  The 
> people who need fast(or at least faster than the current 
> Ruby implementation can provide) performance probably are not using Ruby 
> currently.  I think it's more of a "if you build it, they will come" situation.
> 
Ok, good point.

> I must say that I am a bit worried about the performance issue.  The Perl6 
> folks are working on a new VM that should improve Perl's performance, the 
> Python folks are trying out things like Psyco 
> ( http://homepages.ulb.ac.be/~arigo/psyco/ - Can Ruby benefit from this 
> approach? ) to improve their VM's performance.  I don't see what we're 
> doing in Ruby-land (not that nothing's going on, it's just that I'm not 
> hearing much about it if it is).  We could perhaps get involved in Parrot, 
> but it isn't clear at this point that that would give us the maximum 
>
My take would be a bit different: Why should we do the same thing as
they? Ruby is probably close enough to Python that I'd rather not
try to duplicate their work. Lets do it the Ruby way (if we can find what
that means ;-))!

> possible performance.  There's been a little bit of talk about making a 
> Ruby frontend to SmallScript's VM, but that isn't an open source option. 
> 
That's right.

> You mention that people should pay for faster performance, but I would 
> prefer an open source solution.
> 
Me too, but I'd be willing to pay to get the tools I need; especially if I
decide to go into industry after finishing my PhD and will use Ruby in a
work environment. I'm thinking something like the Aladin license so that
its free if you're free but if you're commercial then pay something (may
not be applicable since they earn money on hardware companys wanting to
include the PS engine in printers etc).

My view is that we can't expect people to develop commercial-grade
software tools on solely their spare time or by using time at other
jobs. Matz got a year or so of funding, and I would guess that something
like that happened to Guido and others in their communities. Or am I wrong
here? I think I'm trying to say that sure if open-source can get us the
tools we want for free then fine, but Ruby feels so important to me I'd be
willing to support its further development.

> Good questions.  I hope they will spark a discussion on improving Ruby's 
> performance (I'm primarily worried about falling behind Perl and Python in 
> this area) that will lead to some solid action.
> 
I hope so too. Not many people have answered though. We'll see...

/Robert