UNIX* System V and 4.1C BSD
From Computer History Wiki
This is a multipart paper posted on usenet comparing 4.1c BSD and Unix SYSV:
Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP Path: utzoo!linus!decvax!harpo!seismo!hao!cires!nbires!ut-ngp!ut-sally!jbc From: j...@ut-sally.UUCP Newsgroups: net.sources Subject: Compare.A Message-ID: <63@ut-sally.UUCP> Date: Sat, 6-Aug-83 22:02:53 EDT Article-I.D.: ut-sally.63 Posted: Sat Aug 6 22:02:53 1983 Date-Received: Sun, 7-Aug-83 17:49:00 EDT
UNIX* System V and 4.1C BSD
John Chambers
Office of Academic Computing & Biostatistics
University of Texas Medical Branch, Galveston
John Quarterman
Computation Center
University of Texas at Austin
{ihnp4,decvax!{eagle,allegra}}!ut-ngp!{jbc,jsq}
{jbc,jsq}@{ut-sally.UUCP,{utexas-11,utexas-780}.ARPA}
Presented at the July 1983 USENIX Conference in Toronto.
cO Copyright 1983 by the Regents of the University of Texas.
ABSTRACT
This paper compares System V (the UNIX system which
Western Electric is currently licensing) and 4.1C BSD (the
final precursor to 4.2BSD, the research UNIX system
developed for DARPA by the University of California at
Berkeley), based on experience with both systems on a DEC
VAX-11/780.
The comparison covers several areas and includes
comments organized by manual section on numerous specific
features (languages, shells, text editing and formatting,
devices, etc.), plus more general and detailed discussions
of such topics as: installation and configuration; sources
and documentation; groups and identifiers; file systems;
interprocess communications; networks; performance
(including some tentative benchmarks); and vendor support.
Common features are mostly left to the manuals, in
order to better concentrate on differences. This is meant
to be a qualitative comparison, intended to serve only as a
guide for further study.
________
* UNIX is a Trademark of Bell Telephone Laboratories, Inc.
CONTENTS
1. Introduction....................................... 2
1.1 Intent....................................... 2
1.2 Format of the Paper.......................... 2
1.3 Disclaimers and Acknowledgments.............. 3
2. Manual Sections.................................... 3
2.1 Commands..................................... 3
2.1.1 User convenience..................... 3
2.1.2 Programming support environments..... 4
2.1.3 Shells............................... 6
2.1.4 Formatting and typesetting........... 6
2.1.5 Graphics............................. 7
2.1.6 Ingres............................... 7
2.1.7 Text editors......................... 7
2.1.8 Electronic mail...................... 8
2.1.9 Printing............................. 8
2.2 System Calls................................. 9
2.2.1 Vfork and fork....................... 9
2.2.2 Reboot............................... 9
2.2.3 Setpgrp.............................. 10
2.2.4 Group system calls................... 10
2.2.5 Ioctls............................... 10
2.2.6 Open and fcntl....................... 10
2.2.7 4.1C BSD file system calls........... 11
2.2.8 Timing............................... 12
2.2.9 IPC.................................. 12
2.3 Libraries and Subroutines.................... 12
2.3.1 Common Object File Format
routines............................. 12
2.3.2 Utmp routines........................ 12
2.3.3 F77 library.......................... 12
2.3.4 Knuth algorithms..................... 12
2.3.5 Software signals and matherr......... 13
2.3.6 Stdio buffering...................... 13
2.3.7 Printf............................... 13
2.3.8 String routines...................... 14
2.3.9 Network library...................... 14
2.4 Devices...................................... 14
2.4.1 Tty.................................. 14
2.4.2 DH-11................................ 14
2.4.3 KMC-11B.............................. 15
2.4.4 VPM.................................. 15
2.4.5 Synchronous terminal................. 15
2.4.6 BLIT................................. 15
2.4.7 Ptys................................. 15
2.4.8 Generalized disk driver.............. 15
2.4.9 Generalized tape driver.............. 16
2.5 File Formats................................. 16
- i -
2.5.1 A.out................................ 16
2.5.2 Ar................................... 16
2.5.3 Fs................................... 17
2.5.4 Termcap and descendants.............. 17
2.6 Games........................................ 17
2.6.1 System V games....................... 17
2.6.2 4.1C BSD ASCII graphics games........ 17
2.6.3 PDP-11 compatibility................. 18
2.7 Miscellany................................... 18
2.7.1 File system hierarchy................ 18
2.8 Maintenance.................................. 18
2.8.1 Init, getty, and login............... 18
2.8.2 Shutdown, halt, and reboot........... 19
2.8.3 Backups.............................. 19
2.8.4 Fsck, fsdb, etc...................... 20
2.8.5 Monitoring and debugging............. 20
2.8.6 Accounting........................... 21
3. Installation and Configuration..................... 21
3.1 Installation................................. 21
3.2 Configuration................................ 22
3.3 Transition................................... 23
4. Sources and Documentation.......................... 23
4.1 Make......................................... 24
4.2 SCCS......................................... 24
4.3 Sources...................................... 24
4.4 Documentation................................ 25
5. Groups and Identifiers............................. 25
5.1 Groups....................................... 25
5.2 Identifiers.................................. 26
6. File Systems....................................... 26
6.1 System V..................................... 27
6.1.1 New file system block size........... 27
6.1.2 Faster access........................ 27
6.2 4.1C BSD..................................... 27
6.2.1 Reimplementation for efficiency...... 27
6.2.2 Other modifications.................. 28
6.2.3 Extended (network) file system....... 28
7. Interprocess Communications (IPC).................. 29
7.1 System V..................................... 29
7.2 4.1C BSD..................................... 29
8. Networks........................................... 30
8.1 System V..................................... 30
8.1.1 X.25................................. 30
8.1.2 PCL network.......................... 30
8.1.3 NSC network.......................... 30
- ii -
8.1.4 RJE to IBM........................... 31
8.2 4.1C BSD..................................... 31
8.2.1 General networking framework......... 31
8.2.2 Variety of hardware and protocols
supported............................ 31
8.2.3 Internet (TCP/IP).................... 32
8.2.4 Berkeley protocols................... 32
8.3 UUCP......................................... 32
8.4 USENET....................................... 33
9. Performance........................................ 33
9.1 Some Qualitative Remarks..................... 33
9.1.1 Paging vs. swapping.................. 33
9.1.2 Terminal I/O......................... 34
9.2 Tentative Benchmarks......................... 34
9.2.1 Load simulation...................... 35
9.2.2 File system throughput............... 36
10. Vendor Support..................................... 36
10.1 Western Electric............................. 36
10.2 U.C. Berkeley................................ 36
10.3 DEC.......................................... 36
10.4 Third Parties................................ 37
10.4.1 OEMs................................. 37
10.4.2 Emulations........................... 37
10.4.3 Consultants.......................... 37
10.4.4 Authors.............................. 38
11. Conclusion......................................... 38
11.1 Selection Criteria........................... 38
11.2 Combinations................................. 38
11.3 Future Directions............................ 39
11.3.1 UNIX standards committee............. 39
11.3.2 Berkeley features and Bell........... 39
11.3.3 Bell licensing and Berkeley.......... 39
Appendix A: Terminology........................... 39
Appendix B: Load Simulation Job................... 41
- iii -
UNIX* System V and 4.1C BSD
John Chambers
Office of Academic Computing & Biostatistics
University of Texas Medical Branch, Galveston
John Quarterman
Computation Center
University of Texas at Austin
{ihnp4,decvax!{eagle,allegra}}!ut-ngp!{jbc,jsq}
{jbc,jsq}@{ut-sally.UUCP,{utexas-11,utexas-780}.ARPA}
Presented at the July 1983 USENIX Conference in Toronto.
cO Copyright 1983 by the Regents of the University of Texas.
ABSTRACT
This paper compares System V1 (the UNIX system which
Western Electric is currently licensing) and 4.1C BSD2 (the
final precursor to 4.2BSD, the research UNIX system
developed for DARPA3 by the University of California at
Berkeley), based on experience with both systems on a DEC
VAX-11/7804.
The comparison covers several areas and includes
comments organized by manual section on numerous specific
features (languages, shells, text editing and formatting,
devices, etc.), plus more general and detailed discussions
__________
* UNIX is a Trademark of Bell Telephone Laboratories, Inc.
1. See Appendix A for the official names of Bell UNIX
Systems.
2. See Appendix A for details about Berkeley Software
Distributions (BSD).
3. Defense Advanced Research Projects Agency (DARPA),
formerly ARPA.
4. VAX, PDP, UNIBUS, MASSBUS, and SBI are Trademarks of
Digital Equipment Corporation (DEC).
- 2 -
of such topics as: installation and configuration; sources
and documentation; groups and identifiers; file systems;
interprocess communications; networks; performance
(including some tentative benchmarks); and vendor support.
Common features are mostly left to the manuals, in
order to better concentrate on differences. This is meant
to be a qualitative comparison, intended to serve only as a
guide for further study.
1. Introduction
1.1 Intent
This paper describes certain differences between
System V and 4.1C BSD, leaving details of common functions
to the manuals. This is a qualitative comparison, intended
to serve only as a guide for further study.
While performance is not a major theme of this paper,
some tentative benchmarks are included to indicate the
relative performance of the two systems. These benchmarks
should not be considered conclusive, since 4.1C is not 4.2
and since we have not had sufficient production experience
with System V.
This paper supersedes a previous paper, ``UNIX
System III and 4.1BSD, A Practical Comparison'', by the same
authors. In some cases, features are noted herein as having
been introduced in System V or 4.1C BSD when they were
actually introduced in System III or 4.1BSD. This usually
occurs when comparisons are being made with V7/32V and is
done simply to decrease the verbiage.
1.2 Format of the Paper
The first section following the Introduction contains
subsections corresponding to sections of the UNIX
Programmer's Manual* in order to provide a framework for
comparison of detailed features of the operating systems.
There follow several sections on subjects which are
wider than a single manual entry or which we consider
important.
__________
* The 4.1C BSD title; see section on Documentation.
- 3 -
Finally, there is a summary section which includes some
comments on recent cooperation among UNIX system developers.
1.3 Disclaimers and Acknowledgments
The authors of this paper are in no way affiliated with
the University of California, Bell Laboratories, or Western
Electric, and are solely responsible for the opinions
presented herein.
4.1C BSD is not a regular Berkeley Software
Distribution and inquiries should not be sent to Berkeley
concerning it. Facilities in 4.1C may be represented
differently in 4.2. In cases in which we know what the
differences will be we have noted them but we do not claim
to have caught every case. When Berkeley is ready to
distribute 4.2BSD, they will announce it.
We would like to acknowledge Dr. Michael Molloy of the
Computer Science Department, University of Texas at Austin,
for the use of the departmental VAX-11/780 and for his
assistance, as well as Bill Lee of the U.T. Austin
Computation Center for his continued moral and material
support. We would also like to thank the following for
reviewing the paper: Sam Leffler of the University of
California at Berkeley, Nina McCloskey of AT&T Technology
and Licensing Division, Armando Stettner of DEC, Dan
Franklin of Bolt, Beranek, and Newman (BBN), and Doug Gwyn
of the U.S. Army Ballistic Research Laboratory (BRL).
2. Manual Sections
The subsections of this section generally follow the
order of the UNIX Programmer's Manual.
2.1 Commands
The general utility commands supplied with the two
systems exhibit relatively minor differences, mainly in
terms of the options available. A few commands are included
in each distribution which do not occur in the other; many
of these are of questionable usefulness anyway and the
reader is referred to the manuals for further details.
Certain larger packages, however, such as language support
facilities, are rather different and are discussed in the
following sections.
2.1.1 User convenience Several utilities are considered
important for the convenience of the frequent user.
- 4 -
Berkeley UNIX provides the page and more file perusal
commands, used to examine a file a screenful at a time. No
equivalent is available in Bell UNIX.
The Berkeley ls command understands proper multicolumn
formatting of a directory listing (when stdout is a tty).
Under System V, ls generates a listing with one entry per
line; a multicolumn listing must be obtained by piping the
output into the paste command, e.g.
ls | paste - - - - -
The Berkeley w program may be used to monitor user
activity; the System V equivalent uses a command file,
/etc/whodo, to generate similar information. However, it is
rather inconvenient to have to specify the absolute pathname
and few users actually have /etc as part of their default
path. (We note, of course, that the superuser's PATH
environment variable does include /etc, perhaps to suggest
that only system administrators and the like should be
interested in such information.)
2.1.2 Programming support environments Several changes
have been made to the C programming support environment
(Software Generation System in WECo parlance) in System V.
Most of the #include files have been rearranged and
expanded, and it is advisable to recompile all C programs.
Pcc, the portable C compiler, includes reasonable
enumerations, changes to structure and union handling
(nonunique structure member names), correct handling of the
void data type, and several bug fixes. The cc command
itself has added the W flag to allow options to be
explicitly specified for a particular compilation subpass.
Certain bugs which are known to remain are documented in the
System Release Description.
Two new tools are included: cxref, which generates
cross-reference listings and obsoletes both cref and xref,
and cflow, which builds a graph of external references
occurring in a collection of assorted source files (C,
assembler, etc.).
The System V f77 programming support environment also
includes two new tools: asa interprets the standard ASA
carriage control characters, and fsplit may be used to split
FORTRAN sources (f77, efl, ratfor) on a procedure-per-file
basis. In addition, the load-time library has been greatly
extended and enhanced.
- 5 -
The libraries for both C and f77 are available in
profiled versions, which must be loaded explicitly in place
of the default, non-profiled ones. These profiled libraries
allow program execution profiling at the library function
level rather than the user program function level.
Further, the symbolic debugger sdb is very much
improved and may be used easily with either C or f77
programs.
The as assembler and ld linker have been modified to
utilize the new Common Object File Format, which is
discussed below. Note that any change to a source file for
a program thus necessitates recompilation of all sources
before the objects may be relinked using ld, since the old
and new object formats are radically different.
The C compiler in 4.1C (pcc) is very similar to the one
in System III, including void, union, enum, and structure
elements named per structure, some of which were added after
32V. Berkeley added very long identifiers in 4.1BSD, while
System III and System V retained the old 7/8 character
identifiers. The as assembler, the ld linker, and
associated libraries are similar to the ones in 32V,
although in 4.1 ld was reworked to be four to five times
faster and this improvement is preserved in 4.1C and 4.2.
The dbx symbolic debugger is new.
4.1C BSD has some bug fixes and other improvements to
f77 (an overlaid version of this compiler is available for
2.8bsd).
4.2 has an extensively reworked version of f77 and its
associated libraries: early versions of this new FORTRAN
package were apparently the source for the new System V
FORTRAN facilities.
Both systems support Ratfor and the Extended FORTRAN
Language (EFL), but 4.1C additionally provides the struct
utility, used to convert FORTRAN sources into reasonably
clean Ratfor.
System V has bs, essentially derived from BASIC. There
is no equivalent in 4.1C BSD; however, the University of
British Columbia BASIC sytem is compatible with 4BSD.
Similarly, System V includes the classic sno SNOBOL
system, while 4.1C includes PASCAL, FRANZ LISP, APL, and fp.
APL is a user contributed software package from Purdue. Fp
(Functional Programming language compiler/interpreter)
implements the applicative language proposed by John Backus
- 6 -
in his Turing award lecture. 4.2 may include Icon as user
contributed software.
There is a COBOL compiler commercially available for
4BSD, and possibly for System V.
2.1.3 Shells System V supports the Bourne shell (sh), with
few noticeable changes from V7. 4.1C BSD has much the same
Bourne shell plus the Cshell (csh), often a new user's first
command language.
The Cshell has most of the capabilities of the Bourne
shell (though the syntax is different), plus the history,
alias and directory stack features. History and alias allow
editing and replaying of saved commands. Such features are
the main reason many users prefer the Cshell (although some
cite its extensive C-like control structures as another
reason).
The 4.1C Cshell also has a set of job control features
(requiring the Berkeley `new tty' terminal driver) which
allow the user to suspend and resume subprocesses.
The 4.1C resource limitation facilities are normally
accessed via the csh limit command. The only close
equivalent in System V sh is the ulimit command, used to
control the size of the file a child process may write.
2.1.4 Formatting and typesetting 4.1C offers the -me macro
package, while System V has the -mm package, somewhat
augmented from PWB. The -ms macros have been removed from
System V but are still found in 4.1C. In 4.2, they have
been extended to provide support for tables of contents and
the like.
System V includes additional macro support for
generating slides and viewgraphs.
An improved interface to the Versatec is provided in
System V, along with new ioctl calls for state control. The
vcat filter for troff which was documented but absent in
System III seems to have disappeared entirely in System V.
Both systems have Versatec drivers expecting a single
interrupt address, whereas the Versatec itself has two
configured into the hardware. 4.1C at least has comments in
the code to tell you this (and #ifdefs to deal with it).
The 4.1 Versatec user programs expected a unit wide
enough to handle four pages abreast; this problem has been
fixed in 4.2 (but not 4.1C) by extensions to the printer
- 7 -
spooling facility.
The Berkeley font library seems rather more extensive
than that provided with System V. These fonts are used by
the Versatec filters to simulate the mounted fonts of a
C/A/T phototypesetter, the standard destination device for
non-device independent troff.
The best version of troff comes with neither of these
systems. This is the Typesetter Independent Troff (TITroff)
package (commonly known as DITroff, for Device Independent
Troff). It is available separately from Western Electric
and includes useful graphics packages (pic and ideal) which
can be used to augment the basic typesetting facilities.
In 4.2, the printer spooling facilities have hooks for
TITroff so that the package can be used immediately when
obtained (though TITroff itself is still distributed
separately by Western Electric). See below under Printing.
The Writer's Workbench facilities style, diction and
explain, which analyze surface characteristics and
readability of written text, are supplied with 4.1C. This
is apparently a Bell Research Group package and is available
separately from Western Electric. Style ignores macros from
-ms, -me, -mm, and -ma, although the manual page only
mentions -ms and -mm.
4.2 also includes an improved refer and bib.
2.1.5 Graphics 4.1C has rather rudimentary graphics
capability.
In contrast, System V has the PWB graphics package,
including ged, a graphical editor, and numerous data
generation, transformation, and display commands. This
graphics capability has been used extensively in conjunction
with the accounting packages.
2.1.6 Ingres The relational database system Ingres is part
of 4.1C, and a commercial version of Ingres is available for
4BSD. We do not know if it will work under System V.
2.1.7 Text editors Both systems have the traditional UNIX
editor, ed, and System V has adopted the Berkeley vi
(screen) and ex (line) editors, which are also of course in
4.1C.
System V documents a new screen editor named se but it
was not included on the distribution. Apparently it does
not utilize the terminal independence capability provided by
- 8 -
termcap but, rather, uses its own terminal description file,
/usr/lib/se.term (also not on the distribution).
Recent versions of the Rand Editor e and UNIX Emacs can
presumably be made to run correctly on System V, although
this was not our experience.
Though the distributed versions of the two commonly
available versions of Emacs have problems running on 4.1C,
since they still attempt to use the obsolete MPX IPC
facility, at least one (Gosling's) has already been adapted
at Berkeley to use the superior Berkeley IPC mechanism. (No
problems were noted running them under 4.1.)
2.1.8 Electronic mail System V has a rudimentary mail
system, not much altered from V7 or System III.
4.1C has a more elaborate one, with most of the
commonly useful mail functions. 4.1C actually provides two
mail delivery routes, one unprotected and the other
encrypted. There is a new mail delivery program called
sendmail (a descendant of the delivermail of 4.1) which
provides a central mail handling system capable of dealing
with multiple networks and addressing formats.
The Rand mh system can be used as an alternative front
end to the Berkeley mail system and will be provided with
4.2 as user contributed software. Some people use Emacs for
this purpose.
We understand that the MMDF mail system from the
University of Maryland can be used with either the Bell or
the Berkeley version of the Unix System but we have no
direct experience with it.
2.1.9 Printing The lpr command and lpd daemon have been
modified in 4.1C to use the file /etc/printcap (similar to
/etc/termcap) to define the characteristics of the various
printers attached to a system. Printers may be added or
deleted without changing the programs and output filter
programs are supported on a per-device basis. It is
possible to treat a printer on another machine as if it were
local (from the user's viewpoint) and have lpd ship files
across a network to it. The Berkeley IPC mechanism is used
for queueing requests, editing the queue, monitoring queue
activity, etc.
In 4.2, lpr, etc., provide support for various raster
devices (such as Varian or Versatec), laser printers (such
as Imagen), and numerous ordinary printers. Specifying a
new type of device in /etc/printcap is relatively easy.
- 9 -
The user specifies a printer either as a command line
option to lpr or in the PRINTER environment variable.
The System V lpr is considered obsolete and has been
replaced by a spooling system similar in flavor to that
provided with 4.1C but without the extensive network
support. The LP-11 is still considered the canonical
printing device, although a particular destination may be
specified by the LPDEST environment variable.
MDQS (Multiple Device Queueing System) is available
from BRL and provides support for queueing output to a
variety of different devices.
2.2 System Calls
The user interface to most of the system calls is the
same, i.e., the interface routines in the C library have the
same calling sequence, but the actual system call numbers
differ.
4.1C has introduced a number of new system calls, some
intended to eventually replace older ones completely. Many
of the older ones are now simulated by interface routines
that call the new, extended ones.
2.2.1 Vfork and fork The fork system call in System V has
been changed to require only one pass through the process
table per invocation. A resulting improvement in performance
is claimed; however, we did not attempt to measure this.
4BSD includes the vfork version of the fork system
call, which allows creation of a new process without the
need for copying the entire address space of the parent.
This makes sense in any environment where processes get very
large, as in the paging environment provided by 4.1C (see
comments below), but the implementation also imposes certain
restrictions which can mislead the unwary. Performance
statistics relating to the use of vfork are widely available
and are outside the domain of this presentation.
2.9BSD has vfork for the PDP-11. 4.3BSD will eliminate
the need for vfork by a reimplementation of fork.
2.2.2 Reboot 4.1C has the reboot system call, which is
quite convenient for persons engaged in system development
work. (See below on the reboot command.)
System V documents a reboot system call for the WECo
3B20S but nothing seems to be available for DEC machines.
- 10 -
2.2.3 Setpgrp 4.1C has elaborated the setpgrp system call
to be more compatible with the job control functions of the
Cshell.
2.2.4 Group system calls 4.1C has a new method of dealing
with the concept of groups and group ids (see the major
section below on Groups).
2.2.5 Ioctls The ioctl system call is essentially
identical in the two systems. The interesting differences
are in the terminal driver ioctls. Both drivers utilize the
``line discipline'' notion, allowing dynamic choice among
several protocols by the user process.
Berkeley offers several new features in 4.1C BSD over
the V7 terminal driver. Some of these are accessed as a new
line discipline (the ``new tty'' discipline), while a few
others are implemented as additional ioctl calls. There is
a line discipline in 4.1C for an RS232 interface to an
Hitachi tablet (this is undocumented). All of these are
useful features, but the tty ioctls have become somewhat
baroque.
The System V terminal driver is radically different
from the V7 one. Many functions which always should have
been orthogonal now are. As one example, the conversion of
carriage return to new line on input and of new line to
carriage return and line feed on output are now separately
controllable functions. Of course, this driver is
hopelessly incompatible with any previous one (except
System III) and with the Berkeley one. Additionally, there
is peripheral processor support for this line discipline in
the KMC-11B (see below).
System V also provides support for a virtual terminal
protocol, allowing drivers for selected terminals to be
compiled directly into the kernel. The terminal type may be
manipulated by two related ioctls, LDSETT and LDGETT; a type
specifier may then be passed to, say, getty (see below).
Unfortunately, this feature is not well-documented and it is
particularly advisable to study the terminal driver code and
the file /usr/include/termio.h.
2.2.6 Open and fcntl The open system call in System V
presents essentially the same interface as in System III but
now claims substantially improved performace due to the use
of a hashed inode table.
The _dup2 function of V7 and 4BSD has been replaced and
elaborated upon in System V by the fcntl system call. 4.1C
preserves the 32V FIOEXCL ioctl call to give control over
- 11 -
the inheritance of file descriptors across an exec; this is
provided by fcntl in Systems III and V. In conjunction with
an additional argument (mode) to the open system call, fcntl
permits access to the O_NDELAY (non-blocking I/O)
capability. (The System III O_NDELAY bug appears to be
fixed in System V.)
4.1C uses an ioctl to set up non-blocking I/O but also
has various open modes in addition to the old read and write
modes, plus the optional third argument for some of them.
Non-blocking opens, for instance, are supported.
4.2 has adopted exactly the same open and fcntl
interfaces as System V, even going so far as to duplicate
the names of the mode bits. A different include file is
used for open, however.
2.2.7 4.1C BSD file system calls 4.1C has a number of new
system calls affecting file I/O, in addition to the
modifications to the open call noted above.
There are now system calls for mkdir, rmdir, and
rename.
Equivalents of old calls that apply to file descriptors
instead of file names have been added: fchown and fchmod.
Symbolic links require some specific calls: lstat,
symlink, readlink.
File truncation is supported by truncate and ftruncate,
and file locking by flock.
Scatter/gather I/O is supported by readv and writev.
The notion of ``file descriptor'' has been generalized
to include various other kinds of descriptors, such as
socket descriptors for use with IPC endpoints. Many of the
system calls, e.g. close, that apply to file descriptors
also have meaning with other types of descriptors, and there
are several new system calls to deal with descriptors, such
as getdtablesize. The most generally useful of the new
descriptor system calls is select, which is used to do
synchronous multiplexing of operations by determining (among
other things) whether it is possible to read or write data
on any of a set of descriptors.
See also the major sections below on File Systems and
IPC.
- 12 -
2.2.8 Timing In 4.1C, all times are returned in a machine
independent format, viz., seconds and microseconds. There
is also improved timezone flexibility. (Systems in
Australia and Europe should no longer experience
difficulties with timezones.)
4.1C uses a simulated 100 Hertz line clock to report
times more accurately than before. The new system call to
replace ftime and time is called gettimeofday. Profiling
using prof is also affected.
The getitimer and setitimer system calls allow the use
of three interval timers, one for real time, one for virtual
time (i.e., the time the process is actually running), and
one for user and system virtual time. The latter allows
interpreters to be profiled, that is, keeping track of when
the interpreted program, rather than the interpreter, is
running. These timers had a bug in 4.1C, but work properly
in 4.2.
2.2.9 IPC The old MPX and FIFO IPC mechanisms have been
largely superseded by the new mechanisms discussed below in
the section on IPC.
2.3 Libraries and Subroutines
There are many changes in this section.
2.3.1 Common Object File Format routines System V adds a
number of routines to provide specialized open, close, seek,
and read operations for files written in the Common Object
File Format (see below).
2.3.2 Utmp routines In accordance with substantial changes
to the format of the utmp structure (see below) in System V,
a collection of routines similar to those provided for
manipulating the password file have been added to deal with
the /etc/utmp and /etc/wtmp files.
2.3.3 F77 library See above under Programming support
environments for f77.
2.3.4 Knuth algorithms In addition to the binary search
(bsearch) and linear search (lsearch) algorithms available
in System III, System V provides routines for searching hash
tables (hsearch) and binary trees (tsearch).
A related UNIX-specific utility, ftw, provides the
ability to recursively descend a directory tree, applying a
user-supplied function at each node. (In other words, it is
the subroutine equivalent of the find command.)
- 13 -
2.3.5 Software signals and matherr System V preserves the
ssignal (and the associated gsignal) facility from
System III, allowing the user to raise and dispose of
software signals.
A related topic is the inclusion in System V of
matherr, an error-handler invoked by functions in the math
library. The user may supply his own version of matherr to
control the disposition of such errors.
4.1C preserves the software signals added in 4.1 to
support the job control features of the Cshell; these are
related to the ``new tty'' line discipline. (4.2 provides a
new signal interface.)
2.3.6 Stdio buffering 4.1C buffers output even if the
output file is a terminal, but flushes all terminal or pipe
output when the process attempts to read from a terminal or
a pipe. 4.1C adapts its buffer size according to the block
size of the file system containing the relevant files, to
produce a marked speed improvement.
The System V stdio buffers have been increased and the
string-oriented output functions have been changed to
provide pseudo-line buffered output to terminals if
buffering has not been specified explicitly. As in 4.1C,
output is flushed on terminal reads.
Both systems keep stderr unbuffered. (The calling
program may, of course, determine the buffering via setbuf.)
In 4.2 perror does a single write (using writev to
gather the arguments) to alleviate the problem of single-
character network transfers, but stderr is still unbuffered.
2.3.7 Printf System V has dropped the System III
(undocumented) vprintf and vfprintf, retaining the V7
(undocumented) doprnt routine, such as is still used in
4.1C. The old (undocumented) %r format has apparently not
been restored, however.
In a similar vein, certain Berkeley programs assume
that sprintf returns the address of the buffer, an
undocumented feature that has been changed and properly
documented in System V (the number of characters written is
returned).
System V printf follows the System III standard, which
abolished the old capital letter formats ("%X", "%F") for
long variables in favor of the prepended-`l' ("%lx", "%lf")
format so that capital letters can be used in the
- 14 -
hexadecimal and floating point formats to mean capital
letters in the output stream.
4.1C has basically the V7 printf and scanf.
The printf and scanf formats are still not fully
compatible in either system.
2.3.8 String routines As in System III, System V changes
the V7 (and 4BSD) index and rindex functions to strchr and
strrchr, respectively, as well as adding a few additional
string routines reminiscent of certain SNOBOL pattern
primitives.
System V also provides new routines to perform basic
memory-to-memory operations (copy, compare, etc.) based on
byte count rather than a terminating null character.
4.1C provides the same functions via the bcopy bcmp,
and bzero routines, which are now in the C library as well
as the kernel.
2.3.9 Network library The 4.1C C library contains a
collection of routines used for translating network-related
names and numbers, such as gethostbyname, which takes the
name of a host and returns its address. There are also a
few routines for manipulating the byte order of network
addresses, such as htonl, which converts a network host
address from network to host byte order, and some routines
brought up from the kernel that are used for manipulating
byte arrays, such as bzero, which clears a byte array.
2.4 Devices
Details of device drivers are beyond the scope of this
paper. We only mention a few corresponding to the most
important devices.
2.4.1 Tty See above under ioctl for a discussion of the
terminal driver changes.
2.4.2 DH-11 System V provides no support for the DH-11
terminal controller. Although DEC no longer supports this
device, many installations either still own DEC DHs or
emulations from other vendors. Also, DEC now supports the
Emulex DH-11 emulation (CS21/H). The replacement is a
combination of DZ-11s controlled by KMC-11Bs. The
System III dh driver is probably portable to System V, but
of course you must acquire a System III distribution.
- 15 -
2.4.3 KMC-11B System V no longer supports the KMC-11A
microprocessor.
The KMC-11B may be used in conjunction with DZ-11s for
offloading terminal I/O processing; it now performs batched
character transfers, an improvement over the character-at-
a-time behavior (a bug) exhibited by System III.
The KMC-11B, as well as the KMS11 (KMC11 plus DMS11-
DA), is also used as the ``Programmable Communications
Device'' (PCD) on which link-level protocols are implemented
under VPM.
2.4.4 VPM The Virtual Protocol Machine (VPM) is a package
which supports a high-level definition language for level 2
protocols to be handled by an interpreter running in a PCD.
In this manner, IBM RJE, a synchronous pseudo-terminal
interface, and several network protocols, including X.25 but
not TCP/IP, may be supported.
2.4.5 Synchronous terminal System V documents support for
a synchronous terminal interface utilizing the Virtual
Protocol Machine, but it was not included in the
distribution.
2.4.6 BLIT At the time of this writing, the driver for the
Teletype 5620 bit-mapped terminal was available only as a
System V-compatible binary object.
2.4.7 Ptys 4.1C has a pseudo-terminal driver to support
network connections. This is actually a driver for two
devices, a slave (/dev/ttyp?) and a master (/dev/ptyp?) end,
where the slave end looks exactly like an ordinary terminal
and the master end (used by network daemons such as rlogind
and telnetd) has a few extra ioctls to aid in simulating a
terminal.
This device is quite important in the common situation
in which there are few directly-connected terminals and most
users log in over a local network.
Pseudo-terminals are also important for such programs
as Emacs and script.
2.4.8 Generalized disk driver System V provides a
generalized driver gd for several moving-head disks (RM05,
RM80, RP04/5/6/7).
The System V driver may be derived from the Berkeley hp
driver, which supports all MASSBUS drives.
- 16 -
The drive type is determined in both systems by
examining the device type register and then using different
parameter tables per drive.
2.4.9 Generalized tape driver A generalized tape driver gt
similar to the generalized disk driver is provided by
System V. This driver offers a general interface to the
TE16 and TU78 style tape drives.
The Berkeley ht drivers support all MASSBUS tape drives
except the TU78, which is supported by the mt driver.
Again, device information is determined by examining
the device hardware type register.
2.5 File Formats
A few file formats are worth mention. In particular,
System V has reorganized several standard file formats, with
important consequences.
2.5.1 A.out The details of the binary object file format
for commands are sufficiently different between the two
systems that it is not possible to run an object file from
one system on the other.
The Common Object File Format (COFF) has been adopted
in System V as the standard output format for programs such
as as and ld in an effort to provide uniformity across
several processors and compatibility with certain other
operating systems. The traditional UNIX a.out header is
included as a part of the COFF header information. Other
COFF information includes such things as the architecture of
the host on which the file was created, line numbers if a
symbolic debugging option was in effect at compile time, and
so forth.
The convert utility may be used to convert pre-System V
objects to COFF.
2.5.2 Ar The System V format for file archives has changed
somewhat from previous releases of UNIX. In particular, an
archive now includes a symbol directory created from the
symbol tables of all archive members which are in COFF to
allow ld to perform random access on the archive. In
addition, numeric data in headers (archive, symbol
directory, member) is stored as 4-byte quantities and should
be portably accessed using the sputl and sgetl library calls
from libld.a.
- 17 -
The portable ar command creates archive headers
specific to the host. Arcv is still available to convert
old-style archives from PDP-11 to VAX-11/780 format.
However, convert should be used to convert pre-System V
archives to the newer format.
4.1C has the ranlib program for inserting an index at
the beginning of an archive, so that the archive can be
randomly accessed.
The Berkeley ar (introduced in 4.1BSD) produces ASCII
output, making the archives portable without the need of any
special libraries.
2.5.3 Fs We consider this topic sufficiently important to
have a major section to itself later in the paper.
2.5.4 Termcap and descendants 4.1C includes the termcap
facility, which maps common terminal control functions to
the specific escape sequences for a particular terminal, and
the curses library of cursor motion optimization functions.
These are used by a number of programs, including the vi
editor, to achieve terminal independence. System V has
adopted termcap but not curses.
Termcap has spawned various lookalikes in 4.1C.
Printcap is used by lpr to determine the characteristics of
printers (see above under Printing). Disktab is used by
newfs to determine how to configure a new format file system
(see below under File Systems). Remote is used by tip (the
successor to cu) to determine the characteristics of a
remote system (see below under UUCP).
2.6 Games
Both systems provide a variety of games, ranging from
the ever-popular hunt-the-wumpus to chess and automated
Dungeons and Dragons.
2.6.1 System V games On System III, most of the games were
shell scripts which echoed the message:
this game does not work on the VAX
This deplorable situation has been largely corrected in
System V.
2.6.2 4.1C BSD ASCII graphics games 4.1C has numerous
games which use termcap and curses to produce ASCII graphics
on various terminals. Examples are rogue (a role-playing
game in the manner of Dungeons and Dragons), worms, rain,
- 18 -
canfield, and mille.
System V has the snake game, but for some reason has
removed the termcap support, making it work on only a few
terminal types.
2.6.3 PDP-11 compatibility 4.1C provides a package which
allows the use of the PDP-11 compatibility feature of the
VAX processor. This package was supposedly included
originally to support the PDP-11 binary of the game dungeon
(zork). The fact that it is still included under games
seems fitting.
2.7 Miscellany
2.7.1 File system hierarchy The most notable addition here
is /usr/ucb, which 4BSD uses to contain objects for commands
developed at Berkeley (though not all such commands), and
/usr/local, used to contain commands and libraries (in
/usr/local/lib) that exist at Berkeley but are not
distributed (making this a convenient place for commands
developed at other sites).
See below under Sources and Documentation for comments
on the source trees.
2.8 Maintenance
4BSD has a reputation in some quarters as being
unsuitable for production use because it is a research
system. This reputation is undeserved, as its maintenance
facilities are more highly developed than those of System V.
2.8.1 Init, getty, and login Init and getty are not
substantially different in 4.1C from 4.1. Login has been
modified to handle rlogin-style remote logins without
passwords on machines under the same administration. There
are also the modifications from 4.1 related to shutdown
(disallowing logins before a shutdown) and security
(prohibiting login by superuser on certain terminals, such
as dialups).
On the other hand, the finite state machine approach
used for System III init has been greatly elaborated in
System V.
Init is driven from the file /etc/inittab, as in
System III. This file is used to specify the identity,
behavior, and arbitrary id of the processes to be associated
with each state init can occupy. A typical entry might
specify that in a state commonly corresponding to multi-user
- 19 -
mode, a getty should be respawned on each terminal line when
the death of a previous getty is detected. Single-user mode
is a distinguished state, with the option of having a
virtual system console connected to any terminal. State
change instructions are issued to the ancestral init via the
telinit command.
The System V getty is likewise driven from a file,
/etc/gettydefs. This file includes, for each speed, initial
and final flags used for setting the mode of a terminal
line, the login herald, and the next speed to try. In
actuality, the speed is only a label for which getty
searches, so that it is possible to make terminal-specific
entries which include control sequences in the login
message, etc.
System V login has mainly been changed to deal with the
new utmp structure. In addition, environment variables may
be set in the login response.
In passing, we note that the old UNIX standby, who, has
been turned into a general utility for summarizing /etc/utmp
and /etc/wtmp. To this end, it now has no fewer than ten
different options.
2.8.2 Shutdown, halt, and reboot 4.1C has the convenient
commands shutdown (bring the system down politely, informing
the users), halt (stop the system immediately), and reboot
(shutdown and bring up a new system).
When coming up, 4.1C automatically performs fsck on all
the file systems (running one fsck subprocess per disk arm,
in parallel, for speed) and brings the system up in multi-
user mode. To bring 4.1C up from a dead start, it is only
necessary to turn the power switch on. (To get into single
user mode, one types ^C or uses another available method.)
The normal method for bringing down System V is to run
the shell procedure shutdown. Other facilities exist for
terminating running processes, including killall, invoked by
shutdown, and fuser, which selectively identifies and kills
processes which are using specified files. A reboot command
is documented for the WECo 3B20S release of System V but
none seems to be available for the VAX. System V has
nothing equivalent to the 4.1C BSD halt command.
2.8.3 Backups 4.1C uses dump for file system backups, in
the V7 manner. The user interface to restor has been
modified, however, to resemble that of tar, making it much
easier to use, as it is now possible to restor by file (or
even directory) name, rather than by inode number.
- 20 -
The 4.1C dump also allows backups over networks. It
runs at tape speed, and is fast enough (especially with a
6250bpi tape drive) that disk-to-disk backups seem
superfluous.
Several backup paths are available under System V. The
volcopy utility from System III may be used to copy a
complete file system. The new finc offers a fast incremental
backup of those files meeting certain selection criteria
(last access, modification, etc.). Frec may be used to
recover files by inumber from a volcopy or finc backup.
Finally, ff, a fast version of find, may be used in
combination with cpio. Dump and restor are not present in
System V.
2.8.4 Fsck, fsdb, etc. A slightly improved version of
fsdb, the interactive file system debugger under System III,
is offered in System V.
Fsck, the file system checker, has been augmented by
dfsck, invoked by checkall, which allows simultaneous file
system checks on two different drives. Note that dfsck
relies on the system being configured to use System V's
multiple physical I/O buffer facility. Also, the use of
dcopy to reorganize the file system for faster access (see
below under File System) will contribute to faster checking.
4.1 added the -p option to fsck, which applies default
rules to preen file systems (usually on reboot), and
incidentally allows concurrent checking of file systems on
different disk arms to speed rebooting. This is retained in
4.1C.
4.1C and 4.2, unlike 4.1, but like System V, requires a
reboot after fsck modifies the root filesystem. In 4.1C and
4.2, unlike System V, the reboot is handled automatically.
2.8.5 Monitoring and debugging System V provides various
facilities inherited from System III for monitoring system
activity and dealing with problems.
An operating system profiling package is available
which uses the pseudo-device /dev/prf to access the
operating system and collect performance statistics by
monitoring selected text addresses.
Extensive error logging and reporting is performed by a
daemon which accesses the /dev/error interface to the system
error collection routines. These reports are often valuable
in analyzing suspected hardware difficulties.
- 21 -
The crash program provides a reasonably clean
interactive utility for debugging core images of the
operating system after a crash. It may also be used to
browse through a running system.
4.1C has syslog to collect kernel error messages into
/usr/adm/messages. Arrangements have also been made to send
many error messages directly to the controlling terminal of
the process that caused them.
There are provisions for analyzing the state of the
paging system after a crash with analyze. There is a paper
on debugging the kernel with adb that tells how to use
numerous canned shell scripts to examine various tables.
Adb itself has the -k option for setting its memory maps
appropriately for the kernel.
2.8.6 Accounting System V provides accounting software
appropriate for a production system in the form of several
tools used to create complex combined reports. The graphics
facilities may be used to automatically produce charts
showing various system parameters (disk reads and writes per
head, number of swaps in and out, kernel buffer statistics,
etc.). These have useful impact in justifying your facility
to upper-level management.
4.1C has kernel hooks to collect similar accounting
information (including paging statistics) but lacks the
graphical output facilities. The facilities provided proved
quite adequate for the purposes of actual system management
in a non-billing environment, however.
3. Installation and Configuration
The installation and configuration documents are
sufficiently complete that few problems should be
encountered when following their instructions. Known
problems are noted below.
3.1 Installation
Both systems are delivered in the traditional Unix
format, viz. a set of half inch magnetic tapes containing
copies of all the binaries, source code, and documentation,
plus accompanying hardcopy documentation (Western Electric
sells manuals ready for use, while Berkeley supplies
duplication-quality masters). 4.2 will come with a console
cassette and floppy, so it will no longer be necessary to
hand-code initial bootstraps.
- 22 -
System V binary licenses are available, and the
Berkeley distribution is also available on two RK07 disk
packs.
For those unfamiliar with VAXen, Installing and
Operating 4.1C BSD contains sections on VAX hardware
terminology and disk formatting which have no counterparts
in Setting Up the UNIX System (the System V installation
guide).
Both systems provide a disk formatter. The format
command provided with System V will format RP06 and RM03/05
disks. The formatter of 4.1C and 4.2 formats almost any
non-DEC UNIBUS or MASSBUS drive, and also includes RM03s and
RM05s, i.e., any disk with the BSC bit in the header. It
cannot handle RP06, RP07, or UDA50s, but the DEC formatter
can do those.
We had no real problems booting either system. (The
System III boot bugs seem to be fixed in System V.)
3.2 Configuration
Both systems are relatively easy to configure.
System V includes driver support for most devices of
interest, including the RM05, RM80, and RP04/5/6/7 disks.
4.1C BSD supports all of the devices just mentioned,
plus many others, and also understands the full
interconnection architecture of the VAX, so that it is
possible to have, say, two RP06's on one MASSBUS and another
on a second, and the system may be permitted to decide at
configuration time which MASSBUS's the RP06's are on.
System V is configured by running the config program
against a system description file; entries in the file are
checked against a list of supported devices in /etc/master.
The vcf command may then be used in standalone mode to
verify address and interrupt information in the kernel
object against the actual hardware present on the system.
4.1C and 4.2 are configured with a config program too,
but this one works markedly differently. The sources are
arranged in such a way that several different kernels (for
the same or different machines) can easily be made from the
same sources. Things such as network node names are
parameterized at run time so that the same kernel can easily
run on several machines with the same CPU and peripherals.
If desired, a generic kernel (such as on the distribution
tape) can be configured that will find likely devices at
- 23 -
startup.
System V still requires hand-setting of numerous
parameters, such as the number of process, file, and inode
table slots, while 4BSD (4.1 and later) decides appropriate
values for these parameters on the basis of one number: the
number of users the machine is to support. (The default
rules can be overridden, of course.)
3.3 Transition
The transition from 4.1BSD to 4.1C BSD (or 4.2) should
not be very troublesome. Though the file system
implementation is quite different, the user interface is
almost identical, especially since system calls that have
been replaced are simulated. The long file names are, of
course, not a problem. The new directory format might
appear to be, but there should be few programs other than
system ones (which are supplied) that read directories
directly.
Directions for the transition are given in
Installing/Operating 4.2bsd under the section ``Upgrading a
4BSD System,'' and A 4.1a User's Guide to 4.1c provides
useful user orientation.
There are no provisions for upgrading from a system
previous to 4BSD, such as 3.0 or 32V, though this could
presumably be done with sufficient investment of effort.
System V is distributed with a document, Transition
Aids, designed to assist the system administrator in
changing from System III to System V.
Especially crucial transition topics include: hardware
support changes (esp. lack of DH-11 support); whether to
convert to a 1K file system; conversion to the new archive
format; and conversion of objects or (preferably)
recompilation of sources for user programs to accomodate the
new headers and object file format.
4. Sources and Documentation
There has been a large amount of reorganization of
sources, documentation, and associated support since 32V.
- 24 -
4.1 Make
System V includes the extended make, which features
many additional default rules to handle common conditions,
to the point that many compilations require no makefile.
Additions are also present which handle archives and SCCS
files (see below) and make use of environment variables and
defaults. Most system programs may be rebuilt by using the
collection of :mk command files located in the source tree.
The 4.1C BSD make seems to be very much in the flavor
of V.7. Rebuilding the whole source tree is as easy as in
System V, however, and is recommended to be done frequently.
4.2 SCCS
System V includes the PWB Source Code Control System
(SCCS), not available in 4.1C BSD. 4.2 is rumored to
include RCS, a public-domain rendering of SCCS.
4.3 Sources
System V preserves the changes to the names of source
directories and files which System III introduced (the
kernel ``sys'' subdirectory becomes ``os'', and ``dev''
becomes ``io''). However, since there is an appropriate
makefile (or :mk command file) for almost everything it is
possible to go to the appropriate parent directory for a
software package and let make do the work.
The 4.1C sources, both user and kernel, have been
radically reorganized in order to simplify recompilation of
the entire system and to promote portability. There is
generally a source directory subtree corresponding to each
directory containing objects, e.g., /usr/src/usr.bin for
/usr/bin, making sources easy to locate. Good use has been
made of symbolic links, in order to avoid duplication of
sources, and to allow keeping certain pieces (such as the
kernel sources) on whatever file system is appropriate,
e.g., /usr/include/sys is a symbolic link to /sys/h, and
/sys is itself a symbolic link to /usr/sys.
The kernel sources have all the VAX-specific code
separated out into different directories and files from the
portable code. The user sources have also been similarly
organized for portability. The C library, in particular,
has been redone. One would expect 4.1C to be as portable to
another 32-bit machine as 32V or System V.
There is a rather widespread problem in Berkeley code
consisting of the use of the type int when long, or even
- 25 -
off_t, or especially time_t is meant. This works fine, as
long as you never try to run such code on a machine where
int is smaller than 32 bits. (This problem is not evident
in the kernel, but rather, in application programs.) This
problem is perhaps less prevalent in 4.1C than in 4.1.
Fairness requires mentioning that there are also
numerous places in the documentation where it is asserted
that int is 32 bits, on the grounds that machines with
smaller word sizes are not sufficient for many of the
functions 4BSD supports.
4.4 Documentation
Berkeley provides the traditional Unix Programmer's
Manual, volumes I, IIA, and IIB, plus an additional volume
of papers written at Berkeley and related directly to the
Berkeley parts of the system. The documents come as
duplication quality masters of 8-1/2 by 11 inch pages
suitable for ordinary three-ring looseleaf binders. The
first volume has of course been updated and is also kept
on-line for easy access.
System V has largely reorganized the system
documentation. Volume one has been divided into a User's
Manual, an Administrator's Manual, and, peripherally, the
Operating System Error Message Manual. Most of the classic
UNIX papers which appear in volume two in the Berkeley
distribution have been pieced together to form such things
as a Document Processing Guide and a Programming Guide. All
in all, there are twelve documents furnished with the
purchase of a System V license; extra copies are for sale by
Western Electric Software Sales and Marketing.
It is disappointing to note that not all of the
documentation is provided on the distribution tape, a
feature considered critical by some (e.g. the sight-
impaired).
5. Groups and Identifiers
4.1C changes the implementation of groups and related
identifiers sufficiently to motivate this section.
5.1 Groups
System V uses the old V7/32V group scheme: a user may
have access to a login group (specified in /etc/passwd) and
also to several other groups (as permitted by /etc/group),
but may be in only one group at a time.
- 26 -
In 4.1C, the same files in the same formats determine
what groups a user is allowed in, but the user is
immediately put in all of them at login: there is no newgrp
command. The groups command lists the groups you are in.
The maximum number of simultaneous groups is a system
compile-time parameter, and the default is eight. The
setgroups system call can be used (by superuser) to set the
groups for a process.
In both systems, each file has a single group
associated with it to determine group read, write, execute,
and setgid permissions. System V creates a new file in the
effective group of the process, whereas 4.1C creates the
file in the group of its parent directory.
Both systems have chgrp (both command and system call)
to change the group of an existing file. 4.1C allows the
user to change the group of a file he owns to be any of the
groups to which he belongs. System V allows the user to
change the user and group id of any file he owns, thus
giving the file away. 4BSD does not, apparently because of
the existence of disk space quotas, which System V lacks.
5.2 Identifiers
Berkeley has extended the setuid and setgid system
calls in 4.1C to allow setting the effective id to the value
of the real id, as well as the reverse. This is very useful
for things like network server daemons, which may now switch
permissions between superuser privileges and those of an
ordinary user, and back, in a single process. This (along
with the socket IPC and non-blocking I/O) allows many
daemons to be implemented as one process where formerly two
were required.
Group ids and process ids are 32 bit integer quantities
in 4.1C. The high order 16 bits of the process id are not
yet used, but probably will be with the development of
distributed applications.
6. File Systems
Both systems have file systems different from their
predecessors and each other. Though the comments in this
section may make the differences seem extreme, the user
interface is not much changed in either case from 32V, and
we have had no trouble transferring files between the two
systems with either tar or cpio (though cpio had to be
ported to 4.1C first, of course).
- 27 -
6.1 System V
6.1.1 New file system block size System V has introduced a
revised file system which allows a choice of a 512 or 1K
byte block. The information concerning the type of a file
system is recorded in its superblock, so it is possible to
have both kinds of file system on the same system.
Robustness is enhanced by carefully controlling the
order in which inode and directory information is written
out in order to prevent serious file system inconsistencies
in the event of a crash.
6.1.2 Faster access Other enhancements claimed to improve
efficiency include multiple (3-7) physical I/O buffers (upon
which dfsck, a multi-drive version of fsck, depends); a
larger number of system buffers (up to 400); free list
management of the file table; and hashing of the in-core
inode table.
A utility, dcopy, is provided to allow reordering of a
file system to optimize access time by compressing directory
``holes'' and spacing file blocks at the disk rotational
gap. Its frequent use is recommended.
6.2 4.1C BSD
6.2.1 Reimplementation for efficiency 4.1C has a file
system that uses a block size and a fragment size that are
settable per file system. The basic block size (usually
4096 or 8192 bytes) is the largest block size used in a
file, and all blocks but the last are this size. The last
one may be any multiple of the fragment size (usually 512 or
1024, and no more than a factor of eight less than the basic
block size). Inodes are divided among several cylinder
groups on a file system, and blocks in a file are usually
localized in a single cylinder group. In-core inode copies
are hashed.
The standard I/O library has been modified to use the
block size returned by the modified stat call to determine
the size of its transfers.
Various changes were made for robustness, as well,
beyond those found in 4.1. For example, static information
from the superblock (such as the block and fragment sizes)
is duplicated in each cylinder group.
Measurements made at Berkeley indicate the new file
system is up to a factor of 16 faster than the old (4.1) one
under ideal conditions, and a factor of 10 is not unusual in
- 28 -
actual use.
4.1C keeps defaults for the various parameters needed
by the new file system in /etc/disktab (a termcap-like
file), where newfs (a frontend to mkfs) uses them in
constructing a file system, storing them in the super block.
Various other programs, such as bad144, which handles bad
sector marking, also use /etc/disktab. This file is a
kludge used because the information is not yet kept on the
disk and accessible by an ioctl.
6.2.2 Other modifications In addition, 4.1C has very long
file names (compile time parameter of 255 characters)
analogous to the long C identifiers, a reworked directory
implementation, symbolic links, and mkdir, rmdir, and rename
as system calls.
The use of file names that are actually 255 characters
long is not, of course, recommended. The idea is to set the
limit high enough that ordinary use will never hit it.
A simulation library for the new directory format has
been distributed several times over USENET; it is a good
idea to use it even if conversion to 4.2 is never planned,
since it solves several old Unix directory access problems
(e.g. insuring null termination of file names extracted
from a directory).
A symbolic link is simply a file containing a pathname,
which is interpreted by the kernel after the pathname of the
link itself. Thus cross-device links and links to
directories are possible.
The motivation for moving mkdir, rmdir, and rename into
the kernel was to make them extensible in a network
environment. In the case of rename, robustness during
system crashes was also a factor.
6.2.3 Extended (network) file system Neither 4.1C nor 4.2
have the extended file system that makes it possible to
mount, on one machine, a file system existing on a disk
connected to another machine, with file transfers then
proceeding over the network connecting the two machines.
There are several implementations of such a facility but
none will appear in 4.2.
- 29 -
7. Interprocess Communications (IPC)
This is one of the areas where the systems diverge the
most.
7.1 System V
System V provides several somewhat different paths for
achieving interprocess communication, mostly developed for
real time support.
The fifo, or named pipe, has been retained from
System III, allowing a process to open a pipe by name rather
than needing a parent to set up appropriate file
descriptors.
The message queue operations associate a unique
identifier with a system message queue and data structure
that includes information about the last processes to send
and receive messages, the times at which these events
occurred, etc.
The semaphore operations associate a unique identifier
with a set of semaphores and a data structure that includes
time and pid of last operation, number of processes
suspended while waiting for a particular change in the
semaphore's value, etc.
The shared memory operations associate a unique
identifier with a shared memory segment (which may be
attached to the data segment of a process) and a data
structure containing the size of the segment, time and pid,
etc.
As an adjunct to the above, process segment locking
(text, data, or both) via the plock system call is also
provided.
The number of message queues, size of each queue,
number of semaphores, number of shared memory segments, etc.
are all parameters which are determined by the system
administrator at system configuration time.
7.2 4.1C BSD
4.1C has dropped the V7 multiplexed files (MPXs) that
were retained in 4.1 in favor of a new interprocess
communication facility. This new socket IPC integrates the
pipes, file and device I/O, and network I/O into one
interface, which allows blocking or non-blocking I/O,
multiplexing several I/O streams in one process by use of
- 30 -
non-blocking I/O and the select system call, and
scatter/gather I/O.
The socket IPC solves most of the traditional Unix IPC
problems, and is more general than the various mechanisms
which have preceded it, such as pipes, MPXs, Rand ports, BBN
await/capac, etc.
The mmap shared memory facility described in the 4.2BSD
System Manual is not supported in 4.1C, and will not be in
4.2. It will, however, appear in 4.3BSD, along with the
revised fork system call that makes vfork obsolete by only
copying pages when they are modified. Various other memory
management-related changes will also come with 4.3.
8. Networks
With the increased use of networks of small
workstations and larger file or compute servers, this
subject is gaining importance.
8.1 System V
While it is said that System VI will incorporate the
Berkeley network code, most network support in System V is
implemented using KMC-11Bs.
8.1.1 X.25 System V documents the use of its VPM facility
to support X.25 in a KMC-11B peripheral processor, and the
same technique can be used for other networks. However, the
X.25 support package was not included on our distribution
tape, and the documentation leads one to believe this was
intentional.
Rumor has it that there is a current project to
implement X.25 under the 4.2 network framework.
8.1.2 PCL network System V provides a driver for the PCL-
11B network bus, used to interconnect multiple CPUs for fast
parallel communications. A local network of UNIX machines
is made practical by the inclusion of the net command, which
allows commands to be executed on remote system. It is very
reminiscent of berknet.
4.2 has a PCL driver.
8.1.3 NSC network System V documents an interface
specification for the NSC A-410 processor and its associated
software, used to access an NSC local net (Hyperchannel).
Neither a driver nor applications software was provided with
- 31 -
the distribution, however.
4.1C and 4.2 have an NSC driver.
8.1.4 RJE to IBM System V implements software which
communicates with IBM JES by emulating a 360 remote
workstation. It relies on a VPM script running in a PCD,
say, the KMC-11B. Facilities are provided for queueing
jobs, monitoring the status of the RJE, and notifying users
of the arrival of output.
8.2 4.1C BSD
Networking is one of the strongest points of 4.1C.
8.2.1 General networking framework The network mechanisms
were designed with the intention of supporting a variety of
network protocols and hardware.
The socket IPC provides an interface common to both
networks (the internet domain in particular) and internal
Unix facilities (the Unix domain).
The internal networking mechanisms support easy
implementation of further protocols or interface drivers,
and are clearly documented.
8.2.2 Variety of hardware and protocols supported Hardware
currently supported includes several kinds of ethernet*
interfaces (3COM, Interlan, Xerox 3Mb experimental), several
ARPANET IMP interfaces (ACC LH/DH, DEC IMP11-A, SRI) a ring
network interface (Proteon 10Mb), and various others, such
as DMC-11, NSC Hyperchannel, and Ungerman-Bass with DR-11/W.
4.2 (but not 4.1C) has a PCL driver.
ISO/OSI** Network, Transport, and lower layer protocols
supported include 3Mb and 10Mb ethernet, Proteon proNET 10Mb
ring, and the DoD internet family (TCP/IP and relatives).
__________
* Ethernet is a trademark of Xerox Corporation.
** International Standards Organization Open Systems
Interconnection: a meta-protocol designed to promote
compatibility among networks.
- 32 -
8.2.3 Internet (TCP/IP) Both TCP and UDP are available for
use with IP either on the local network or over the
internet. ICMP is supported, and there are some gateway
facilities.
The socket IPC, together with a network library,
provides many of the functions of the Session layer. The
socket type SOCK_STREAM, which provides a reliable, ordered,
byte stream, is currently supported by TCP, while
SOCK_DGRAM, providing datagram service, is supported by UDP.
There is no internet protocol to support SOCK_SEQPACKET, for
sequenced packets. SOCK_RAW allows direct access to, for
instance, the IMP interface, for debugging and development
of new protocols. Only the superuser is permitted to use
this socket type.
At the Applications layer, the Internet protocols
Telnet, FTP, SMTP, and TFTP are supported.
8.2.4 Berkeley protocols The berknet facilities of 4.1 are
officially removed from 4.1C and 4.2.
There are also various new protocols developed at
Berkeley, including remote login among machines under the
same administration without passwords (rlogin/rlogind).
Remote shells and remote procedure calls (courierd) are
supported, as are file copy (rcp/rshd), and status protocols
such as the one that supports rwho and ruptime (rwhod).
This latter takes advantage of the broadcast packet facility
of Ethernets and rings to exchange status information about
who is on what system and what systems are up on the local
network. (The idea is easily extensible to networks without
broadcast packets.) Some of these protocols use UDP and some
use TCP.
These protocols make use of several machines over a
local network or networks quite convenient.
8.3 UUCP
Both systems support UUCP, though the details diverge:
System V allows uucp copy addresses to specify paths across
multiple systems (as for mail), while 4.1C still permits
copies only between adjacent systems. Naturally, all
systems in a multisystem path must be running uucp with
forwarding to properly effect forwarding.
4.2 has all well-known uucp bugs fixed. It also
supports more than half a dozen auto-dialers. The spooling
directories have been made sane.
- 33 -
The cu command is retained in System V, but dropped by
4.1C in favor of tip. Tip gets parameters for a remote
system from /etc/remote (yet another termcap-like file) so
it is possible to just type
tip system-name
and be connected to the remote system (whose name, system-
name, is in /etc/remote), without having to know the phone
number or devices involved.
If the cu interface is desired, linking tip to cu is
all that is required to get it. Various cu-like commands
are supported directly by tip.
System V includes a library routine, dial, used to
establish a dialout connection. This routine is used by cu,
but, curiously, uucico still relies on the same old conn.c
code.
8.4 USENET
Neither system includes USENET news program sources.
4.2 will provide both USENET programs (readnews, postnews,
et al.) and notesfiles, as user contributed software.
In any case, the best version of news is clearly Bnews
2.10, which has been distributed over USENET. One method to
get it might be to set up a UUCP connection to a neighboring
USENET site and copy it over.
The USENET and UUCP networks have become widespread
enough that connection to them is certainly beneficial.
9. Performance
This is a sticky issue which we will not treat in
detail, as this is not a performance evaluation
presentation. We will give a few tentative benchmarks, and
mention two qualitative performance areas.
9.1 Some Qualitative Remarks
Two important areas where performance varies widely due
to system configuration and usage are paging vs. swapping
and terminal I/O.
9.1.1 Paging vs. swapping System V, like System III, 32V,
V7, and all the PDP-11 Unixes, swaps, while 4.1C, like 4.1,
pages. With enough memory on a VAX-11/780, it is difficult
- 34 -
to tell the difference for a load of small processes,
because System V just doesn't swap.
If it is desirable to run huge graphics processes or
many Emacs editors or the like, the telling point is not so
much the performance as the virtual address space provided
by the 4.1C paging system. Such things as LISP require the
large address space paging provides, and Ingres is much more
usable with it, since it can run as one process instead of
half a dozen.
We certainly do not intend to indicate, however, that
we think paging and swapping produce equivalent performance.
There are many technical papers on comparative performance
that indicate paging gives much better performance; it is
merely that our (admittedly idiosyncratic) experience was
that under a light load it is hard to tell the difference
without measuring it.
9.1.2 Terminal I/O Using DH11 terminal controllers, 4.1C
provides reasonable terminal I/O performance. Berkeley has
modified the DZ11 driver sufficiently that even these
(basically interrupt per character) devices are usable. It
should be remembered that DEC does not provide DH11
controllers for VAXen. This affects DEC maintenance, though
similar hardware is available from other vendors.
If you need numerous terminals running at 9600 baud or
higher, the System V combination of DZ11s and KMC11 terminal
controllers seems preferable. The other side of this coin
is that little choice has been left for System V users,
since DH-11 driver support is not included in the
distribution and since DZs alone are unlikely to yield
acceptable response.
9.2 Tentative Benchmarks
These measurements were taken on a VAX-11/780 with six
megabytes of memory and a single RP07 disk.
The disk was partitioned into three sections which had
similar sizes under the two operating systems. The various
kernel parameters were chosen by configuring 4.1C for 32
users, and, for System V, by picking the largest parameter
values suggested in the documentation. The resulting
numbers of buffers, inode and file structures, etc., were
similar. Memory was interleaved.
No particular care was taken beyond these steps to tune
either system to its maximum performance.
- 35 -
The numbers given here should not taken too literally,
but only as indicative.
9.2.1 Load simulation This was done using a program that
forks ptys instances of itself, and then each pty* repeats a
job forever, or rather, until the run is over, as decided by
the original parent program. The job is a brief shell
script that uses commands common to both systems, as found
on each system. Sources to compile with cc were taken from
System V to avoid the long filename problem. They were
carefully picked to avoid any System V peculiarities, such
as getopt. Files to use with nroff were also taken from
System V, for no particular reason.
The output was redirected to a file to avoid terminal
I/O considerations.
Repeated runs were taken until the job throughput
stopped decreasing due to file system degradation. The
following figures were obtained:
ptys 1 2 4 8 16
System V
jobs/hour/pty 46 25 13 6.4 3.2
jobs/hour 46 50 51 51 51
4.1C BSD
jobs/hour/pty 60 32 16 8 4
jobs/hour 60 64 64 64 64
The total number of jobs per hour increased slightly
from one to two ptys, and then remained constant, as all CPU
cycles were absorbed. The number of jobs per pty is, of
course, just the total divided by the number of ptys.
We interpret these results to mean that 4.1C is
noticeably faster than System V. We do not state the
obvious figure of 25%, because the results could easily be
varied by, for instance, increasing the amount of file I/O a
job uses (to take advantage of the faster 4.1C file system),
__________
* No pty devices were used: this term is used only for
convenience.
- 36 -
or by using larger processes (to force System V to swap,
which it never did with the above job).
Tuning either kernel could, of course, vary the results
either way.
Definitive benchmarks will have to await the release of
4.2BSD.
9.2.2 File system throughput Our experience has been that
the 4.1C filesystem is markedly faster than the System V
one. However, the actual figures vary so much according to
the size of the files used, the transfer block size, the
filesystem block size, whether memory is interleaved or not,
etc. (though under all conditions we have tried 4.1C is
faster than System V), that it would take some months and
another paper the size of this one to deal with the problem.
Rather than present partial and possibly misleading figures,
we have decided to not present any.
10. Vendor Support
The amount and variety of support for UNIX has
increased dramatically over the last few years.
10.1 Western Electric
Western Electric supports System V on VAXes and the
larger PDP-11s, providing software assistance and user
training. (User training is now available, though software
assistance has apparently not yet been fully implemented.)
10.2 U.C. Berkeley
The University of California at Berkeley cannot,
because of the nature of the institution and of the funding
used to support the development of Berkeley UNIX, provide
commercial support. They do, however, accept bug reports,
which may affect future versions. See below on DEC.
10.3 DEC
Digital Equipment Corporation has announced they will
support UNIX in the manner they have supported VMS as a VAX
operating system (and they will also support it on PDP-11s
(V7M)). This is apparently basically Berkeley UNIX, i.e.,
4BSD (and 2BSD).
- 37 -
10.4 Third Parties
The number of organizations dealing with UNIX these
days is quite large.
10.4.1 OEMs Many companies bringing out new Motorola
68000-based systems recently have chosen System III as the
base for their operating system, with the apparent intention
of moving to System V. To some extent, this will no doubt
lock them into System V, and persons wanting to buy
something close to a small turnkey system will probably wind
up with essentially Bell UNIX.
Other manufacturers with microprocessors likely
targeted for System V ports are Intel, National, and Zilog.
There are several ports of 4.1 to the 68000, and at
least one of 4.2. There are also at least two ports of 4.1
to the National Semiconductor 16032.
Several of the vendors offering System III based 68000
systems claim to support ``Berkeley enhancements,'' the
interpretation of which varies between vendors, but usually
seems to include vi, ex, termcap, and curses, and sometimes
more.
10.4.2 Emulations Several emulations of UNIX are available
from third parties, either software vendors or universities.
Typically these are designed to provide a UNIX environment
on top of another operating system, generally VMS. The
quality of emulation varies from implementation to
implementation, as does the concept of what ``UNIX'' should
look like.
On a slightly different note, a package will be
available from BRL in the very near future which emulates
System V on top of 4.2BSD.
10.4.3 Consultants There is a new class of companies that
produce neither hardware nor software but instead provide
assistance in obtaining and supporting both. These mostly
try to cater to the markets for both systems.
There is a large amount of free software available for
4.1 (and thus 4.1C) that was written principally at academic
institutions. Much of it is portable to System V, though
something like Interlisp that requires a huge address space
is not, and there are problems with many things like Emacs
because of the use of long identifiers.
- 38 -
Most commercial vendors attempt to produce and sell
software packages to run on either variety of UNIX. Bell is
among these vendors, with the TITroff package, the S
statistical package, etc.
Many of the commercial vendors using System III
(System V) have produced graphical, menu-driven interfaces
for the naive user, so that it is never necessary to deal
directly with any UNIX shell. These mostly require bit-map
terminals, varieties of which are also available from other
vendors.
The famous Bell Blit bitmap terminal is available from
Teletype (model 5620). Unfortunately, as noted previously,
the Unix software is available only as a System V binary.
10.4.4 Authors A number of books designed to assist the
new UNIX user have recently appeared.
Most of these either attempt to steer a neutral course
by describing what is essentially V7, making them less
useful in either a 4.2 or System V context, or they closely
follow System III (V) in hopes of describing what will come
to be a ``standard''.
The 4.1C (4.2) user is left with the traditional task
of reading the manuals.
11. Conclusion
A brief summary may be useful.
11.1 Selection Criteria
One may choose either Berkeley or Bell Unix on the
basis of a particular needed function, such as network
support, because of performance in one area or another,
because of the support of a particular vendor, or for some
other reason. We have touched on all these areas above, we
hope in sufficient detail to indicate the capabilities of
the two systems, so that areas for further investigation
will be clear.
11.2 Combinations
For companies with the resources, the best solution is
probably to run either 4.1C BSD or System V and port the
desired facilities of the other. This is the traditional
route. An alternative is the aforementioned package from BRL
or something similar.
- 39 -
Even companies with no desire to merge the two systems
would be well-advised to get some sort of expert support
(whether in-house or not), as neither Bell nor Berkeley can
be counted on to offer the really broad support
traditionally supplied by hardware vendors for their
operating systems. This situation may change in the case of
System V as more sites begin running the system and
demanding the support which has been promised, but at the
moment only time will tell. The same applies to DEC's
support of 4BSD.
11.3 Future Directions
A few recent developments may indicate a trend away
from continued fragmentation of the UNIX community, and
especially from the divergence of the systems offered by
Berkeley and Bell.
11.3.1 UNIX standards committee The /usr/group UNIX
standards committee appears to be making progress in
standardizing at least the most basic facilities of the
operating system, and has representatives from most segments
of the community.
11.3.2 Berkeley features and Bell The inclusion of vi, ex,
and termcap in System V, as well as the adoption of a 1Kbyte
block file system, shows that Bell is aware of the work
Berkeley has been doing for years in researching new
directions. Perhaps System VI will go further and adopt,
for instance, csh, and paging.
11.3.3 Bell licensing and Berkeley Unfortunately, until
recently it has not been possible for Berkeley to include
software from Bell licenses later than 32V, because the
price would have been prohibitive for many of the Berkeley
licensees. Though the recent reform of Western Electric's
licensing scheme apparently came too late to affect 4.2BSD,
perhaps we will see Berkeley adopt some later-day Bell
developments.
Appendix A: Terminology
The official names of the various versions of the Unix
System developed by Bell Laboratories and previously or
currently available from Western Electric are:
o+ UNIX Time-Sharing System, Sixth Edition (V6);
o+ UNIX Programmer's Work Bench (PWB), V6 plus SCCS, etc.;
- 40 -
o+ UNIX Time-Sharing System, Seventh Edition (V7), the
PDP-11 version of the first portable UNIX system;
o+ UNIX/32V Time-Sharing System Version 1.0 (32V), like
V7, but for the VAX;
o+ UNIX System III (System III), combining PWB, V7, and
32V;
o+ UNIX System V (System V), now being licensed.
There have been numerous Berkeley Software
Distributions of the various Berkeley versions of the Unix
System.
o+ 2BSD is used herein as a generic term for the PDP-11
distributions.
o+ 2.8BSD is the latest PDP-11 distribution in general
use.
o+ 2.81BSD was a an intermediate system that was never
officially distributed, but is in use at several
ARPANET sites with a port of the 4.1A network software
incorporated into it.
o+ 2.9BSD is the distribution just now being licensed, and
is said to make a PDP-11 look like a VAX 4BSD system.
o+ 3.0BSD was the first paging system for the VAX, derived
from 32V.
o+ 4.0BSD was the second Berkeley VAX distribution.
o+ 4BSD is used herein as a generic term for any Berkeley
VAX distribution from 4.0BSD on.
o+ 4.1BSD is the VAX distribution in most common use, and
contains numerous improvements over 4.0BSD.
o+ 4.1A BSD, 4.1B BSD, 4.1C BSD were versions intermediate
between 4.1 and 4.2. None of them were available
outside of Berkeley except for beta test, and none of
them can be ordered from Berkeley.
o+ 4.2BSD will presumably be licensed soon.
- 41 -
Appendix B: Load Simulation Job
This is contents of the shell file that was used in the
load simulation:
mkdir $1; cd $1
cc -o simple -p ../simple.c
simple
nroff -man ../prof.1
prof simple
tar -cvf /dev/null ../simple.c simple mon.out
rm simple mon.out
nroff -man ../termio.7
cc -o cmp ../cmp.c
cd ..
rm -rf $1
