From: Eero Saynatkari <ruby-ml@...>
Date: 2009-02-16T08:32:52+09:00
Subject: [ruby-core:22171] Re: A short memorandum (Re: 1.8.7 Specifics)

Excerpts from Headius: Charles Oliver Nutter's message of Sun Feb 15 23:02:05 +0200 2009:
> Hongli Lai wrote:
> > I understand that this is frustrating to you as a Ruby implementer, but 
> > I still don't think that it's a bad idea to expose GC APIs. The GC APIs 
> > can be documented as "private" or "not guaranteed to be available on 
> > other Ruby implementations" or something, so that when people file bug 
> > reports, you can just point to the documentation and tell them that it's 
> > their responsibility to fix their apps.
> > For example, Phusion Passenger is fully compatible with standard MRI and 
> > only uses the COW GC API when available. If things break then we 
> > consider that to be a bug.
> 
> I never said don't expose them...just don't make them part of the 
> always-visible set of APIs. Attempting to load an extension for access 
> to the APIs and failing over is much easier than going one by one 
> through the individual methods and checking if they exist.

I would prefer having to `require "gc"` for the insulation. Only
applying to the API, of course: I see no reason to require people
to specifically activate the GC functionality itself.

On the other hand, the GC module already exists: should it be
moved out of core altogether?


--
Magic is insufficiently advanced technology.