.ce
VT11 DRIVER HISTORY

The first HUJI VT11 driver was based on a system-buffers resident
display file. D_file and light-pen information were accessed
via read-write-seek system calls, and
the buffers' busy flag was used as a d_file access semaphor.

Being extremely slow (nevertheless, working and bugless!)
it was replaced by a low-core-shared-data-area based driver.
D_file was now inside the user's address space, and light-pen
information was moved, by system interrupt routines,
to a structure residing at the beginning of d_file.
This gave as much real-time response as any UNIX process can.
Timing was enables by 'vtsync' system-call (wait for GT
interrupt).
This scheme was very useful
and at this time GLIB (interactive graphics package) was written,
but the
shared-data-area jazz did not work smoothly
with some UNIX features (e.g. break system-call),
and d_file loops without an interrupt command or
a runaway GT sometime caused users yo sleep forever
on 'vtsync'.

When a PDP-11/10 arrived and the VT11 was moved there
(making a GT42),
a stand-alone system was written, simulating part of the
UNIX environment. This was done so that user programs
using Glib did not have to be modified (not all of UNIX
system calls are supported, of course).
