In article <199802131302.AA13698@world.std.com> you write:
> code a better text terminal and include that in kernel space along with 
> the video bit maps so memory protects are not violated.  I'm sure doing 
> in kernel space has problems as it would be fairly large as it would also 
> have to include the character maps for at lease a basic font.  There are 
> details I may be missing but it's nto trivial as there would be some point 
> where the boot console is effectively gone and the system console takes 
> over.  That may upset the OS as the the error log goes there and and would 
> be "missing" for a period of time while the replacement console takes 
> over.

It isn't that hard to deal with.  Many platforms NetBSD runs on have a
similar situation (actually, which don't? pc532 and mvme68k are the only
ones I can think of, off the top off my head).  The pmax and (optionally)
sparc ports use "rcons" (raster console).  Amiga and several other m68k
ports use variations of "ite", etc.  Most of them do fairly reasonable
emulations of VTxxx.  There is also a (slow) movement towards a generic
MI framework for workstation consoles called "wscons" to replace all of
these.