louk@tslwat.UUCP (Lou Kates) (11/09/90)
The following describes the relative advantages and disadvantages
of SCCS vs. RCS. The * indicates which of the two on each line is
preferable. In lines without a *, there is either no difference
or preference will depend on your particular needs and
circumstances.
Item SCCS RCS
custom header yes* no
change requests yes* no
make understands SCCS* does not understand RCS
setuid yes* no
user interface worse better*
symbolic tags no yes*
loc'n of revs current directory any directory*
rev history stored in SCCS can be copied to file on checkout*
speed retrieve old versions faster retrieve latest version faster
distribution binary from UNIX vendor source from Purdue or GNU
free front ends sccs CVS
usually used in System V BSD
The above list may change over time as SCCS and RCS are improved.
In particular, apparently, setuid on certain platforms and the
ability to retrieve older versions faster will be in the next or
future versions of RCS.
The disadvantages with either can often be overcome by using a
simple shell (possibly even just a csh or ksh alias) front end.
Several freely available front ends exist.
Perhaps one of the points in the table above is enough to make
you choose one of the two systems but my sense of the response
was that RCS and SCCS are sufficiently similar that you would
probably be guided by what was most available and familiar to
you which means SCCS on System V and RCS on BSD.
WHERE TO GET
RCS is available on arthur.cs.purdue.edu. If you have problems
contact rcs-bugs@cs.purdue.edu. An MS-DOS version is available on
wsmr-simtel20.army.mil in
PD1:<MSDOS.PGMUTL>DOSRCSEX.ZIP
PD1:<MSDOS.PGMUTL>DOSRCSMP.ZIP
PD1:<MSDOS.PGMUTL>MKSRCSEX.ZIP
On star.cs.vu.nl in pub/cvs is a shell script version of the CVS
front end for RCS by Dick Grune, Vrije Universiteit, de Boelelaan
1081, 1081 HV Amsterdam, the Netherlands, dick@cs.vu.nl.
On uunet.uu.net in ~ftp/comp.sources.unix/volume22/cvs-berliner
is a C version of CVS front end for RCS by Brian Berliner, Rocky
Mountain Technology Center, Sun Microsystems, 5465 Mark Dabling
Blvd., Colorado Springs, CO 80918, berliner@sun.com.
RCS and CVS may also be found on prep.ai.mit.edu although the
locations referred to above may have more recent versions.
A number of SCCS front ends have been posted to comp.sources.unix
and alt.sources.
REFERENCES
Marc J. Rochkind, "The Source Code Control System",
IEEE Transactions on Software Engineering, Vol SE-1(4), 364-370, Dec 1975.
Walter F. Tichy, "RCS -- A System for Version Control",
Software--Practice and Experience, Vol 15(7), 637-654, Jul 1985.
THANKS
Thanks to the following for posting or sending me emails:
bad@atrain.sw.stratus.com (Bruce Dumes)
berliner@rmtc.Central.Sun.COM (Brian Berliner)
bhahn@bogus.sw.stratus.com (Bill Hahn)
cbmvax!hutch!purdom@uunet.UU.NET (purdom)
celit!billd@celerity.fps.com (Bill Davidson)
charles@phx.mcd.mot.com (Charles Wolff)
dick@cs.vu.nl (Dick Grune)
drz@csri.toronto.edu (Jerry Zarycky)
emv@math.lsa.umich.edu (Edward Vielmetti)
guy@auspex.auspex.com (Guy Harris)
jmm@eci386.uucp (John Macdonald)
johnl@iecc.cambridge.ma.us (John R. Levine)
leadsv!white@apple.com (Ben White)
mks!watmath!domain.com!mdv (Mike Verstegen)
rick@tetrauk.tetrauk (Rick Jones)
rob@b15.INGR.COM (Rob Lemley)
sdrc!scjones%thor@uunet.UU.NET (Larry Jones)
watmath!acf4.NYU.EDU!poppers (Michael Poppers)
watmath!servio.SLC.COM!penneyj (D. Jason Penney)
woods@eci386.uucp (Greg A. Woods)rickert@mp.cs.niu.edu (Neil Rickert) (11/09/90)
In article <2252@megadon.UUCP> louk@tslwat.UUCP (Lou Kates) writes: >The following describes the relative advantages and disadvantages >of SCCS vs. RCS. The * indicates which of the two on each line is >(...) > >Item SCCS RCS > >make understands SCCS* does not understand RCS This may be version dependent. The 'make' I am using understands RCS. I am not sure whether it understands SCCS. (On Umax 4.3, 'make' is as supplied by the vendor, Encore). -- =*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*= Neil W. Rickert, Computer Science <rickert@cs.niu.edu> Northern Illinois Univ. DeKalb, IL 60115. +1-815-753-6940
heiney@wsl.dec.com (Bob Heiney) (11/09/90)
The Free Software Foundation's make program includes support for RCS, so you can have automatic check out if you want to. Bob
bruce@balilly.UUCP (Bruce Lilly) (11/15/90)
>The Free Software Foundation's make program includes >support for RCS, so you can have automatic check out >if you want to. Last time I looked, GNU make didn't support SCCS however. Specifically, GNU make doesn't handle the ~-hack in the SUFFIX rules for identifying SCCS files. If one has many makefiles which are designed to handle SCCS files, there is strong incentive to not use GNU make. On the other hand, if one uses a ``standard'' make, there is no support for a ',' as a delimiter for suffixes, so it's rather difficult to automate use of RCS. The result: migration from one system to the other, or use of both SCCS and RCS can be difficult. -- Bruce Lilly blilly!balilly!bruce@sonyd1.Broadcast.Sony.COM
guy@auspex.auspex.com (Guy Harris) (11/15/90)
>make understands SCCS* does not understand RCS There do exist versions of "make" that understand RCS, although the vanilla S5 one doesn't have built-in RCS support the way it has built-in SCCS support, and I don't think it's easy to add in without source. The SunOS 4.x "make" doesn't have built-in RCS support the way it has built-in SCCS support, either, but you can add rules to help. There also exist versions of "make" that don't understand SCCS.... >speed retrieve old versions faster retrieve latest version faster WRONG. Older versions are NOT faster in SCCS. Didn't I say that in my message?
cudcv@cu.warwick.ac.uk (Rob McMahon) (11/21/90)
>On the other hand, if one uses a ``standard'' make, there is no support for a >',' as a delimiter for suffixes, so it's rather difficult to automate use of >RCS. Uh ? You don't need it. Most of what you need can be done with rules like .SUFFIXES: .c,v .c,v.c: $(CO) -q $*.c .c,v.o: $(CO) -q $*.c $(CC) $(CFLAGS) -c $*.c $(RM) -f $*.c ... etc. Note that for SCCS it *has* to be built into make, because it uses prefixes instead of suffixes. Rob -- UUCP: ...!mcsun!ukc!warwick!cudcv PHONE: +44 203 523037 JANET: cudcv@uk.ac.warwick INET: cudcv@warwick.ac.uk Rob McMahon, Computing Services, Warwick University, Coventry CV4 7AL, England
donn@hpfcdc.fc.hp.com (Donn Terry) (11/21/90)
Item SCCS RCS
speed retrieve old versions faster retrieve latest version faster
This isn't quite true:
SCCS: Linear in the size of the SCCS file for any revision.
(Retrieve any version in the same amount of time.)
RCS: O(n), where n is the number of changes between the top and the
requested version. (Retrieve old versions very slowly.)
In practice, top revision is slightly faster (for the same history)
for RCS, but small enough in either case as to not really matter (a
few seconds at most; not enough time to think about the difference).
Donn Terry
HP, Ft. Collins.balilly@Broadcast.Sony.COM (Bruce Lilly) (11/28/90)
>>On the other hand, if one uses a ``standard'' make, there is no support for a >>',' as a delimiter for suffixes, so it's rather difficult to automate use of >>RCS. > >Uh ? You don't need it. Most of what you need can be done with rules like > >.SUFFIXES: .c,v > >.c,v.c: > $(CO) -q $*.c >[ ... ] >... etc. Note that for SCCS it *has* to be built into make, because it uses >prefixes instead of suffixes. The premise of Rob's ``solution'' is that *all* files have a '.' somewhere. This doesn't work, e.g., for s.Makefile -> Makefile. For that to work, a ',' would have to operate as a SUFFIX delimiter, which it does not do in ``standard'' make. There's no way to make make make Makefile from Makefile,v using ony SUFFIX-based rules. An explicit dependency line and rule would be required for each and every file. P.S. Note that make will automatically do a get if there is an s.Makefile (or s.makefile) and neither Makefile nor makefile exist. In short, make and SCCS are well integrated. -- Bruce Lilly blilly!balilly!bruce@sonyd1.Broadcast.Sony.COM
eggert@twinsun.UUCP (Paul Eggert) (12/07/90)
There's no way to make Makefile from Makefile,v using only SUFFIX-based
rules. An explicit dependency line and rule would be required for each
and every [such] file.
Not true. Here's how to do it:
.SUFFIXES : ,v
,v :
co -q $*
Lilly also writes:
make will automatically do a get if there is an s.Makefile (or s.makefile)
and neither Makefile nor makefile exist. In short, make and SCCS are well
integrated.
make and RCS are similarly well integrated, if you have a modern version of
make, e.g. GNU make 3.58. For older makes, you can use the trick described
above.woods@eci386.UUCP (Greg A. Woods) (12/07/90)
In article <2282@megadon.UUCP> balilly@Broadcast.Sony.COM (Bruce Lilly) writes: >[....] > The premise of Rob's ``solution'' is that *all* files have a '.' > somewhere. This doesn't work, e.g., for s.Makefile -> Makefile. For that > to work, a ',' would have to operate as a SUFFIX delimiter, which it does > not do in ``standard'' make. There's no way to make make make Makefile > from Makefile,v using ony SUFFIX-based rules. An explicit dependency line > and rule would be required for each and every file. [ I *really* like your use of the English language! ;-) ] You can always use something like the .sh rule by calling your file "Makefile.sh". Non-traditional, and potentially confusing, but it works. > P.S. Note that make will automatically do a get if there is an s.Makefile > (or s.makefile) and neither Makefile nor makefile exist. In short, make and > SCCS are well integrated. Unfortunately I find make a little lacking when it comes to either RCS or SCCS. Gnumake is slightly better, and perhaps dmake is too, but they are non-standard, and versions I've tried were either buggy, or not compatible with SysV make. SysV make won't find SCCS/s.[Mm]akefile, for example. Instead, I prefer to "keep it simple, stupid", and not deal with the version control through make. I use RCS or SCCS with "sccs" in a mode where they are transparent, and make need know nothing about them. All files are always "checked out" and visible to make. The very nature of the way one uses "sccs", for example, ensures this is always true. -- Greg A. Woods woods@{eci386,gate,robohack,ontmoh,tmsoft}.UUCP ECI and UniForum Canada +1-416-443-1734 [h] +1-416-595-5425 [w] VE3TCP Toronto, Ontario CANADA "Political speech and writing are largely the defense of the indefensible"-ORWELL
jfh@rpp386.cactus.org (John F Haugh II) (12/27/90)
>make and RCS are similarly well integrated, if you have a modern version of >make, e.g. GNU make 3.58. For older makes, you can use the trick described >above. Except for the initial checkout of Makefile. I have two versions of a collection of software I manage in SCCS right now. The version is selected by a flag contained in the Makefile. If there is no currently checked out version, it gets the latest one. "Standard" make won't do this if RCS is used. I'd love to run GNU software, but I just don't have the disk space to spare ;-) -- John F. Haugh II UUCP: ...!cs.utexas.edu!rpp386!jfh Ma Bell: (512) 832-8832 Domain: jfh@rpp386.cactus.org