Mr. B. E. Carpenter
MPS Division
CERN
1211 Gene`ve 23
Switzerland

Dear Mr. Carpenter:

This is in response to your letter of November 27, reference
MPS/CO/BC/je.
The UNIX system, its software, and its documentation
are proprietary to the Bell System and are currently being
released, under a license from the Western Electric Company,
only to educational institutions.
The possibility of extending such licenses to governmental
and commercial organizations is being considered, however.
To obtain information about such a license, you may write to

	Mr R. G. Shahpazian
	Room 400
	Western Electric Company
	222 Broadway
	New York, New York, 10038

You should also send a copy of your letter to

	Mr. S. P. Morgan
	...

Mr. Shahpazian will acknowledge your letter and will
inform you as soon as a decision has been reached.

In answer to your specific questions:

There is currently no user-available method to lock a process
in core to meet real-time constraints.  This should
be very easy to add, however, since the system
has internal to it a flag which prevents swapping the
process's core image.
Somewhat more work would be required to arrange that
physical core occupied by the process lies at the
edge of the available space, which is desirable to prevent
fragmentation of the physical address space.

The system will run with only one RK disk, although naturally
both response time and available disk space are better with the
swapping disk space and user file space on
separate devices.
Incidentally, I think the SIGOPS conference version
of the paper was overly optimistic about the
amount of memory needed;
we are unable to support machines with less than
32K words of core.

There is no automatic means of initiating a process in response to
an interrupt.
The approach we would take to this need is to have an existing
process which wakens in response to the interrupt and creates
a new process.  Unless the time requirements are very strict
I don't think this is a problem.

Although UNIX for non-segmentation PDP-11's still exists,
we are unable to support it, since its internal structure is 
so different from the current system and since
it requires so much handcrafting to adapt it to the
hardware configuration.

It is trivial to introduce a new user program such as a
compiler, since one need merely place the file
containing the program in a particular directory.
New device drivers are in general fairly easy,
since the interface between the system and the drivers
is quite clean.  It does require some
knowledge of the way the system operates.

Fortran definitely supports the floating point hardware;
in fact we recommend it if much use is to be made of Fortran.
The compiler generates the same kind of semi-interpretive
"threaded code" as does the DEC Fortran; I suspect it
is somewhat slower than DEC's since Fortran is not heavily
used by us, and we have not felt impelled to optimize
as carefully.
If you are at all adventurous in this area I can advise
using "C", which is the language in which
UNIX itself and most of its software are written.
C is a true compiler which generates excellent code and
supports both single- and double-precision floating-point
arithmetic.

I can give no definite answer as far as cost is concerned.
As I indicated, UNIX is
currently being made available only to educational
institutions; this is on a cost-free basis.
I can't predict any charges that might be
made should licenses be extended
on a broader basis, but I doubt if they would seem excessive.
I'm not sure what you mean by "software support."
Of course all the ordinary software comes with the system:
editor, language processors, utilities, and so forth.
If you refer to continuing support,
the situation is still not very clear.
We hope to be able to send out occasional mailings
describing bugs found and new software, and to
generate tapes of the appropriate source programs.
Still, you will have to take into account
the fact that your only source
of information is a 2-man operation an ocean away.

			DMR

Copy to SPM
