[Openmcl-devel] Universal binaries? Closed source applications?

[Paul Onions]

On 24 Jul 2010, at 01:08, Gary Byers wrote:

> The issue that I've always gotten stuck on has to do with how a  
> universal
> kernel finds an architecture-specific heap image.  If no command- 
> line arguments
> override this behavior. the kernel looks for a file whose name  
> matches its own
> with the string ".image" appended, and it had always seemed that  
> that meant
> that we'd need some sort of support for fat heap image files.

Okay, but this raises a couple more questions in my mind, the first  
being how do we create a universal kernel? The second question is, do  
we need access to both Intel and PPC Macs to create the architecture- 
specific image files?

> The LLGPL (under which CCL is licensed; see ccl/doc/LICENSE) tries  
> to clarify
> what this means in the context of a lisp application, where the  
> lines between
> "the library" and "the application" are blurrier.  The general idea  
> is the same:
> you don't need to make the sources to your application available in  
> order to
> be compliant,  if you change CCL itself, you do need to make the  
> sources of
> those changes available.  Adding a method to a predefined generic  
> function or
> subclassing a predefined class isn't considered to be a change to  
> the "library";
> changing the implementation of a predefined method or function is,  
> and it's
> the nature of lisp that one could probably think of things that're  
> less clear-cut.
> (In a lot of clear-cut cases - not all, but a lot - it's probably in  
> your interest
> to make the sources to a change to CCL itself public, to help ensure  
> that future
> changes to CCL don't break them.)

Thanks for your clarification of the LLGPL in this situation. It's  
basically what I thought it would be.

Thanks very much for your help,
Paul