From: Brian Mitchell <binary42@...>
Date: 2010-01-15T03:05:42+09:00
Subject: [ruby-core:27590] Re: better GC?

2010/1/14 Gon�alo Silva <goncalossilva@gmail.com>:
> So, we're getting the worse of the two options. GC will still be slow and
> people will need to fix their ruby gems. I don't see how this is good for us
> or Ruby itself.

The key is that there would be a correct way to write things that
could avoid most of the penalty, if I read that right. You'd pay a tax
for extensions that aren't using the new API. So they are slow, what's
new? The option here is to give us some future and the ability to
improve the software we do have control over.

Assuming we don't jump to the ideal GC approach in one big bang...
What we need to decide is what the most appropriate first steps are. I
think there is room to employ new APIs. Penalties are fair. If this
weren't the case, then someone might be complaining that their 286
running in real mode should be able to run a modern operating system
right now. We had to take a hit with that software compatibility and
even performance as protected modes were added.

Let's not freeze progress because something someone did might need
rework. For those who don't want the penalty, well, they can keep
using the current implementation as that's probably what they would be
stuck without these proposed efforts anyway.

Brian.