rouben@math13.math.umbc.edu (09/03/90)
In a previous article Michael Meissner writes: >> How common are sh's that don't understand shell functions? > > Anything that is BSD based without adding from System V. For example, > the Ultrix DECstation that I'm posting from has the musty BSD shell as > /bin/sh and the System V.2 shell with shell functions as /bin/sh5. > Another problem with the BSD shell Ultrix uses, is that test (aka '[') > is a separate command that you have to fork/exec to get to -- I > suspect echo is too..... In spite of its shortcommings, ultrix's BSD sh beats sh5 in terms of efficiency. On a DECstation5000, we have: ls -l /bin/sh /usr/bin/sh5 -rwxr-xr-x 1 root system 45056 Apr 1 12:27 /bin/sh -rwxr-xr-x 2 root system 163840 Apr 1 09:56 /usr/bin/sh5 So the executable for sh5 is more than three times larger than that of sh. For quick execution of simple tasks there is a sense of overkill in using sh5. I guess that's the reason that the folks at DEC have not replaced sh with sh5. -- Rouben Rostamian Telephone: (301) 455-2458 Department of Mathematics and Statistics e-mail: University of Maryland Baltimore County rostamian@umbc.bitnet Baltimore, MD 21228, U.S.A. rostamian@umbc3.umbc.edu
gt0178a@prism.gatech.EDU (BURNS,JIM) (09/03/90)
in article <3865@umbc3.UMBC.EDU>, rouben@math13.math.umbc.edu says: > -rwxr-xr-x 1 root system 45056 Apr 1 12:27 /bin/sh > -rwxr-xr-x 2 root system 163840 Apr 1 09:56 /usr/bin/sh5 Well, that's amusing. Try: -rwxr-xr-x 1 root 23552 Nov 19 1987 /bin/sh* -rwxr-xr-x 1 root 39936 Nov 18 1987 /usr/bin/sh5* on Ultrix V2.2 Worksystem 1.1 (I forget what model CPU). -- BURNS,JIM Georgia Institute of Technology, Box 30178, Atlanta Georgia, 30332 uucp: ...!{decvax,hplabs,ncar,purdue,rutgers}!gatech!prism!gt0178a Internet: gt0178a@prism.gatech.edu
guy@auspex.auspex.com (Guy Harris) (09/04/90)
>So the executable for sh5 is more than three times larger than that >of sh. For quick execution of simple tasks there is a sense of >overkill in using sh5. I guess that's the reason that the folks at >DEC have not replaced sh with sh5. I tend to be leery of guesses, perhaps because there may be more than one possible answer to the question to which you're guessing the answer. One alternative answer is that they're afraid putting in an S5 shell - even with changes to make it more V7/BSD-flavored, such as changing the behavior of "echo" - might break some V7/BSD shell scripts. Sun took the alternate tack, in SunOS 3.0 (yes, 3.0, not 3.2; it wasn't done as part of the big S5 compatibility push, it was done to add in the functionality and bug fixes of the S5R2 shell), by making the aforementioned changes and making an S5R2-based shell "/bin/sh". (The change in 3.2 was to make it act as if it were picking up builtins such as "echo" and "test"/"[" as programs, based on PATH, so that the behavior of those builtins depended on whether "/usr/5bin" came before "/bin" or "/usr/bin" in PATH.) For comparison, here are the sizes of the 4.3BSD shell source, compiled on a Sun-4 running 4.0.3 (after fixing the "old-style" initializations that the SunOS 4.x C compilers no longer accept) and "/usr/bin/sh". Both are statically linked, and both are compiled with "-O": auspex% size ./sh /usr/bin/sh text data bss dec hex 40960 8192 0 49152 c000 ./sh 90112 8192 3160 101464 18c58 /usr/bin/sh so the code slightly more than doubles in size, but not triples. I have no problem with doubling the size of the code, if it means I don't have to write "least common denominator" shell scripts and don't have to stick something gross and OS-dependent such as "#! /usr/bin/sh5" at the front of the script.
meissner@osf.org (Michael Meissner) (09/04/90)
In article <3865@umbc3.UMBC.EDU> rouben@math13.math.umbc.edu writes: | In a previous article Michael Meissner writes: | >> How common are sh's that don't understand shell functions? | > | > Anything that is BSD based without adding from System V. For example, | > the Ultrix DECstation that I'm posting from has the musty BSD shell as | > /bin/sh and the System V.2 shell with shell functions as /bin/sh5. | > Another problem with the BSD shell Ultrix uses, is that test (aka '[') | > is a separate command that you have to fork/exec to get to -- I | > suspect echo is too..... | | In spite of its shortcommings, ultrix's BSD sh beats sh5 in terms of | efficiency. On a DECstation5000, we have: | ls -l /bin/sh /usr/bin/sh5 | | -rwxr-xr-x 1 root system 45056 Apr 1 12:27 /bin/sh | -rwxr-xr-x 2 root system 163840 Apr 1 09:56 /usr/bin/sh5 | | So the executable for sh5 is more than three times larger than that | of sh. For quick execution of simple tasks there is a sense of | overkill in using sh5. I guess that's the reason that the folks at | DEC have not replaced sh with sh5. Just about every 'simple' shell script that I have uses test and/or echo. If there are any savings in the V7 shell, they just got lost by having to do a fork/exec combination. Also, I remember that when V.2 came out, people measured the 'new' shell with the V.1 shell which was more or less the V7 shell, and found by doing a reimplementation, it made the 'new' shell much faster. I don't remember the numbers, but it was something like 20% faster, even after factoring out the builtins. I finally sovled the problem Guy Harris mentioned in another article (ie, do you gunk up your shell scripts with #! /bin/sh5 and make them completely non-portable), by noticing that for the machines I care about, the Korn Shell was installed in /usr/bin, so I now tend to use #! /us/bin/ksh, which is still a kludge. I happen to think that System V integration was real low priority on Ultrix. All you have to do is look at how they botched curses (requiring you to change the source to include cursesX.h instead of curses.h). Sun did a much better job of mixing the two. -- Michael Meissner email: meissner@osf.org phone: 617-621-8861 Open Software Foundation, 11 Cambridge Center, Cambridge, MA, 02142 Do apple growers tell their kids money doesn't grow on bushes?
vixie@wrl.dec.com (Paul Vixie) (09/04/90)
In article <MEISSNER.90Sep3231422@osf.osf.org>,
meissner@osf.org (Michael Meissner) writes...
# I happen to think that System V integration was real low priority on
# Ultrix. All you have to do is look at how they botched curses
# (requiring you to change the source to include cursesX.h instead of
# curses.h). Sun did a much better job of mixing the two.
As far as that goes, HP did an even better job than Sun. And Pyramid
did a better job that ANYBODY. But why does anybody care about System V?
(Not speaking for DEC -- we seem to care about System V...)
--
Paul Vixie
DEC Western Research Lab <vixie@wrl.dec.com>
Palo Alto, California ...!decwrl!vixiewnp@iiasa.AT (wolf paul) (09/04/90)
In article <3865@umbc3.UMBC.EDU> rouben@math13.math.umbc.edu writes: >-rwxr-xr-x 1 root system 45056 Apr 1 12:27 /bin/sh >-rwxr-xr-x 2 root system 163840 Apr 1 09:56 /usr/bin/sh5 > >So the executable for sh5 is more than three times larger than that >of sh. For quick execution of simple tasks there is a sense of >overkill in using sh5. I guess that's the reason that the folks at >DEC have not replaced sh with sh5. Huh? On our VAX6210 running Ultrix 3.1: -rwxr-xr-x 1 root 23552 Jan 11 1989 /bin/sh -rwxr-xr-x 1 root 44032 Jan 11 1989 /bin/sh5 That's not even twice the size! I suspect that on your machine, sh is stripped, and sh5 is not. And SOME increase in size is almost necessary for an increase in functionality :-) ... -- Wolf N. Paul, IIASA, A - 2361 Laxenburg, Austria, Europe PHONE: +43-2236-71521-465 FAX: +43-2236-71313 UUCP: uunet!iiasa.at!wnp INTERNET: wnp%iiasa.at@uunet.uu.net BITNET: tuvie!iiasa!wnp@awiuni01.BITNET * * * * Kurt Waldheim for President (of Mars, of course!) * * * *
meissner@osf.org (Michael Meissner) (09/04/90)
In article <1990Sep4.063729.27822@wrl.dec.com> vixie@wrl.dec.com (Paul Vixie) writes: | In article <MEISSNER.90Sep3231422@osf.osf.org>, | meissner@osf.org (Michael Meissner) writes... | # I happen to think that System V integration was real low priority on | # Ultrix. All you have to do is look at how they botched curses | # (requiring you to change the source to include cursesX.h instead of | # curses.h). Sun did a much better job of mixing the two. | | As far as that goes, HP did an even better job than Sun. And Pyramid | did a better job that ANYBODY. But why does anybody care about System V? | | (Not speaking for DEC -- we seem to care about System V...) Because each of the various UNIX camps have good things to offer (such as the V.2 /bin/sh and curses in the System V case). IMHO, I think Pyramid's solution (separate att/ucb universes) is the wrong solution. I want a merged system, not two systems under the same roof...... -- Michael Meissner email: meissner@osf.org phone: 617-621-8861 Open Software Foundation, 11 Cambridge Center, Cambridge, MA, 02142 Do apple growers tell their kids money doesn't grow on bushes?
dd26+@andrew.cmu.edu (Douglas F. DeJulio) (09/05/90)
wnp@iiasa.AT (wolf paul) writes: > In article <3865@umbc3.UMBC.EDU> rouben@math13.math.umbc.edu writes: > >-rwxr-xr-x 1 root system 45056 Apr 1 12:27 /bin/sh > >-rwxr-xr-x 2 root system 163840 Apr 1 09:56 /usr/bin/sh5 > > -rwxr-xr-x 1 root 23552 Jan 11 1989 /bin/sh > -rwxr-xr-x 1 root 44032 Jan 11 1989 /bin/sh5 > That's not even twice the size! > > I suspect that on your machine, sh is stripped, and sh5 is not. I suspect that the first figures are for a DECstation and the last are for a VAXstation. Remember, RISCish binaries will almost always be much bigger than CISCish binaries. -- DdJ
terryl@sail.LABS.TEK.COM (09/05/90)
In article <13289@hydra.gatech.EDU> gt0178a@prism.gatech.EDU (BURNS,JIM) writes: >in article <3865@umbc3.UMBC.EDU>, rouben@math13.math.umbc.edu says: >> -rwxr-xr-x 1 root system 45056 Apr 1 12:27 /bin/sh >> -rwxr-xr-x 2 root system 163840 Apr 1 09:56 /usr/bin/sh5 > >Well, that's amusing. Try: > >-rwxr-xr-x 1 root 23552 Nov 19 1987 /bin/sh* >-rwxr-xr-x 1 root 39936 Nov 18 1987 /usr/bin/sh5* ls -l is pretty useless on executables. What if one executable had its symbol table stripped, and the other one didn't???? Try size(1) instead. It has more meaningful output.... text data bss dec hex 90112 8192 3112 101416 18c28 /bin/sh.bsd 92176 8528 7652 108356 1a744 /bin/shs.sysv And here's the ls -l output: -r-xr-xr-x 1 bin root 101216 Jun 20 15:25 /bin/sh.sysv -rwxr-xr-x 1 root other 133120 Feb 20 1990 /bin/sh.bsd Note that the bsd sh is actually smaller according to size(1), but ls -l shows it is bigger, due to the symbol table.....
gt0178a@prism.gatech.EDU (BURNS,JIM) (09/05/90)
in article <MEISSNER.90Sep4112609@osf.osf.org>, meissner@osf.org (Michael Meissner) says: > IMHO, I think Pyramid's solution (separate att/ucb universes) is the > wrong solution. I want a merged system, not two systems under the > same roof...... Sequent Dynix did the same thing, and the ATT universe is CRIPPLED *if* you try to use it by itself. However, I've been adding the ATT bin directories to my UCB universe path, and this works fine for all pre-compiled ATT universe commands - including very SysV specific things like ipcs/ipcrm to manage shared mem. & other IPCs. The only time I need to switch to the ATT universe (usually just by prefixing the current command with ATT, tho' I could go interactive) is when I want 'cc' to link in ATT libraries and include files, or I want to look at SysV versions of the man pages. (For some reason ATT man doesn't use 'more' - it's not even in the ATT bin directories.) -- BURNS,JIM Georgia Institute of Technology, Box 30178, Atlanta Georgia, 30332 uucp: ...!{decvax,hplabs,ncar,purdue,rutgers}!gatech!prism!gt0178a Internet: gt0178a@prism.gatech.edu
jand@kuling.UUCP (Jan Dj{rv) (09/05/90)
In article <1990Sep4.063729.27822@wrl.dec.com> vixie@wrl.dec.com (Paul Vixie) writes:
#In article <MEISSNER.90Sep3231422@osf.osf.org>,
#meissner@osf.org (Michael Meissner) writes...
## I happen to think that System V integration was real low priority on
## Ultrix. All you have to do is look at how they botched curses
## (requiring you to change the source to include cursesX.h instead of
## curses.h). Sun did a much better job of mixing the two.
#
#As far as that goes, HP did an even better job than Sun. And Pyramid
#did a better job that ANYBODY. But why does anybody care about System V?
#
Well, HP actually started with System V and mixed in BSD...
Jan D.chet@cwns1.CWRU.EDU (Chet Ramey) (09/06/90)
In article <1649@kuling.UUCP> jand@kuling.UUCP (Jan Dj{rv) writes: >Well, HP actually started with System V and mixed in BSD... I don't think that this is true. Way down deep inside the kernel, HP-UX is really a descendent of 4.2 BSD. Most of the user-level code is from System V, though. Chet -- Chet Ramey ``Levi Stubbs' tears run down Network Services Group his face...'' Case Western Reserve University chet@ins.CWRU.Edu