MiniUnix/usr/doc/new/memo3

Find at most related files.
including files from this version of Unix. 
.NH
System Features
.PP
The Operating System itself is written
in the high-level language, C (2) and as
such bears a strong resemblance to the standard
UNIX Operating System which runs on the
PDP-11/40, 11/45 and 11/70 computers.
Because of the memory address space limitation
due to the lack of a segmentation unit,
the system size is generally kept to
12K words in size.
Thus its capabilities are somewhat less
than those of the standard UNIX system,
especially in the area of inter-process communication
and interactive debugging.
The number of processes (and users) which MINI-UNIX (MX)
can comfortably support is in general much less than for the
standard UNIX system.
No more than four users are recommended for MX.
Currently the number of processes allowed is thirteen.
This is a 'sysgen' parameter.
The address space available to a user program is 16K words.
This enables almost all user programs which run under Version 6 UNIX
to run unmodified under MX.
The exceptions are noted in a later section.
The capabilities of MX fall somewhere in between those
of standard Version 6 UNIX and those of the LSX
operating system (3).
LSX is a single-user UNIX-compatible system
for the LSI-11 micro-processor using
floppy disks as secondary storage.
.PP
MX supports all of the UNIX system calls of the Version 6
UNIX operating system with the exception of:
ptrace, pipe, prof, setgid and getgid.
For the sake of completeness the status of implementation
of the system calls are summarized in an Appendix.
Thus MX supports all UNIX user programs with the exception
of programs which use the above mentioned system calls.
User programs are compiled and relocated to start
at address 060000 and may occupy up to 16K words of memory.
Up to 13 processes may exist at any one time, although only one process
may be in core and running.
This should be sufficient to handle up to four simultaneous users.
Since no memory relocation is available, the complete
user program image must be swapped out to bring in
a new user program.
Hence no sharing of text is possible.
No software memory management is required.
Scheduling is done on a simple round-robin basis
with each process in the run state receiving
a two second time slice.
.PP
The file system supported by MX is identical to that provided
by the UNIX time-sharing system.
The structure of the super-block and of the file inodes
is maintained.
MX supports the identical file system hierarchical
structure and makes the same distinctions between
ordinary files, directories and special files.
Removable file systems are supported as well.
Hence 'mount' and 'umount' system requests are treated
identical to those in standard UNIX.
Under MX, a file's size is limited to one megabyte.
Large files are supported, but huge files (two level
indirect block) are not.
Groups are not supported in MX.
Thus a file has only a given owner user ID and no group ID.
Read, write and execute permission bits
are available for both owner and non-owner of a file.
The set-user ID bit is also supported in MINI-UNIX.
.PP
No interactive debugging is possible in MX since the 'ptrace'
system call is not implemented.
One may still use the C debugger 'cdb' but cannot plant
breakpoint traps in the running image of a child process.
It may only be used for post-mortem debugging on
core images.
The profiling of a process to determine where
it is spending its time is also not permitted.
Pipes have not been implemented in the MX system
for two reasons.
One is that it requires too much code in the system
which is already butting its head
against the top.
Two is that the overhead required in switching
from the process writing the pipe to the process
reading the pipe involves a process swap.
Filters, however are simulated at the command level
as discussed in the next section.
.PP
Some other features have been stripped out of the system
in MX in the interest of minimizing system address space.
Upon reading a file block, "read-ahead" is not invoked.
This contributes only slightly to a loss in throughput.
Physical I/O has also not been implemented.
This precludes the ability to read large contiguous
pieces of disk directly into the user's address space
without system side-buffering.
The system buffering scheme is much simpler than in standard UNIX
but a maximum of only eight buffers may be allocated
in a minimum system configuration.