> It has also come to my attention that Wolfgang has his solution working
> already, so why no have him commit it to the tree? :-) I haven't seen
> any messages really opposing his proposal after he properly explained
> it to us :-) Talked about this enough, let's get on and improve our
> operating system :-)

Well, it's not in the tree, since:

1. It currently works only with a.out format.

2. It keeps the symbol table in main memory which is inelegant to say the least.

The main problem with making the symbols pageable is that the current VM system
doesn't allow anything to give it additional pages after its initialization.
Since the in-core symbol table can be copied to VM and be swapped only after
the VM is initialized, its pages cannot be used for anything else anyway.
The necessary changes to the VM code shouldn't be too hard, at least not in
the MACHINE_NONCONTIG case (which is essentially code stolen from Mach 3.x, and
Mach 3.x does allow later addition of pages).

BTW, what port(s) still uses the old non-MACHINE_NONCONTIG VM model?  I really
would like to see this dropped.
--
ws@TooLs.DE     (Wolfgang Solfrank, TooLs GmbH) 	+49-228-985800