2022-10-18 12:26:01 <KipIngram   > I'm contemplating a new approach to my Forth development.
2022-10-18 12:26:46 <KipIngram   > The idea would be to do it all from Python.  Generate the ELF file directly.  I currently use a Python script to write an include file with a bunch of my primitives - this would extend that idea "all the way."
2022-10-18 12:27:29 <KipIngram   > This would be for getting it to the point where it could load the remainder of the system itself, from disk blocks.
2022-10-18 12:28:11 <KipIngram   > I feel like if I do that right, I'd be in a better position to make fairly global changes to the system design.
2022-10-18 12:29:00 <KipIngram   > And I'd LIKE to do it so that there's a clean layer that defines the processor interface, so I could target x86 or ARM or whatever by adjusting that layer only.
2022-10-18 12:31:10 <KipIngram   > I'd also like to wind up with a system where the "system" lives as a shared resource, that can be used by any thread.  Starting a new thread housing a new system would add almost no space initially.  Sort of a "green threads" philosophy.
2022-10-18 12:55:02 <KipIngram   > Being able to have a shared system probably means NOT using offset-based stuff like I have my last couple of rounds.  When NEXT picks a CFA up from a definition, it won't have any way of distinguishing a system word CFA from a local word CFA.  It needs to just be an absolute address so that that it does the same thing regardless.
2022-10-18 13:22:59 <KipIngram   > I could bypass the ELF format altogether, though, and plan on using a "loader program" to allocate RAM, put my code in it, and jump to it.  That might give me more flexibility.
2022-10-18 13:23:49 <KipIngram   > The loader might also be able to feed source to the newly loaded process, until it reaches its ability to load source itself.