toma@attctc.Dallas.TX.US (Tom Armistead) (01/04/90)
I have a "requirement" to provide hard facts regarding COBOL performance
under UNIX. I am working for a client that is building a Unix system and
they only have COBOL programmers. They will not accept my word on the
fact that C is a better performer and easier to use (in UNIX). Has anyone
here ever seen any benchmarks between C and COBOL under UNIX or does anyone
know of PRINTED material that specifically states why COBOL is worse than
C on unix systems....
I feel like I'm talking to a tree in trying to convince these people to
use C for development. Any help would be greatly appreciated!!!
P.S. This is for a Store Automation project (~Point of Sale~).
Thanks,
Tom
--
-------------
Tom Armistead
UUCP: {ames,lll-winken,mit-eddie,osu-cis,texbell}!attctc!swsrv1!tomarwa@cs.AthabascaU.CA (Ross Alexander) (01/05/90)
toma@attctc.Dallas.TX.US (Tom Armistead) writes: >I have a "requirement" to provide hard facts regarding COBOL performance >under UNIX. I am working for a client that is building a Unix system and I like C. I _love_ C, and do all my work in it. I liked B before I liked C (yes, B is a real language and a precursor to C), and I liked assembler-H and GMAP before I liked B or C. I hate COBOL, and hated it from day one. My personal biases are well formed and clear. HOWEVER: there is no reason on earth (modulo lazy compiler and library implementors) why COBOL performance should be particularly bad, especially on a machine as inherently CISCy as a '386. Developement may be slower, debugging may be a little awkward ( "MODIFY dweeb TO PROCEED TO looser." indeed :-P ), but to slap screens up, manipulate ISAM files, and do decimal arithmetic, COBOL is just fine. And COBOL coders come cheaper than C hackers. Besides, how much performance does POS require? It's inherently rate limited. Humans can only shove stuff through a checkout so fast. Getting things so that you can carry 20 registers rather than 15 per server is nice, granted, but the % marginal improvement in overall system price is probably going to be small. And the labour involved in getting C to do mixed-precision fixed decimal arithmetic is going to eat that advantage anyway. Not to mention the maintenance headaches. I once wrote a POS application for LAN'ed 8088's; yes, it got written in C. I _think_ I needed the performance (also, I didn't happen to have a COBOL developement system). You have 20 or 30 times the horsepower, perhaps "efficiency" redefines itself in such an environment. And another besides: with luck, the silly b*stards will go broke ;-) -- -- Ross Alexander (403) 675 6311 rwa@aungbad.AthabascaU.CA VE6PDQ
rogerk@sco.COM (Roger Knopf 5502) (01/05/90)
In article <10821@attctc.Dallas.TX.US> toma@attctc.Dallas.TX.US (Tom Armistead) writes: >I have a "requirement" to provide hard facts regarding COBOL performance >under UNIX. I am working for a client that is building a Unix system and >they only have COBOL programmers. They will not accept my word on the >fact that C is a better performer and easier to use (in UNIX). Has anyone I don't accept that whole, either. While C is undoubtedly faster for almost all applications, it is not necessarily easier to use, especially if you have only COBOL programmers on staff. Also, C may be faster but COBOL is no slouch if it is a good compiler and there are excellent compilers available for Unix. >here ever seen any benchmarks between C and COBOL under UNIX or does anyone >know of PRINTED material that specifically states why COBOL is worse than >C on unix systems.... > >I feel like I'm talking to a tree in trying to convince these people to >use C for development. Any help would be greatly appreciated!!! C is way overblown as an applications development language. In general it always requires additional software - a file manager and a screen manager - to write commercial applications. COBOL has both. While I would not necessarily choose COBOL first, or even over C, the fact that your client already knows COBOL makes it a good choice. Roger Knopf SCO Consulting Services If caught, my employer will disavow all knowledge of me or my actions....
johnl@esegue.segue.boston.ma.us (John R. Levine) (01/06/90)
toma@attctc.Dallas.TX.US (Tom Armistead) writes: >I have a "requirement" to provide hard facts regarding COBOL performance >under UNIX. ... It would be interesting to see some Cobol performance figures for a 386. The 386 should be a pretty good Cobol machine -- it has instructions to support packed and unpacked decimal arithmetic as well as all sorts of string munging, which I'd think are the main sorts of computations that Cobol programs do. PC Cobol implementations have tended to be compile-and- interpret since much of what a Cobol program does would be subroutine calls anyway, and I don't know how that affects performance in practice. -- John R. Levine, Segue Software, POB 349, Cambridge MA 02238, +1 617 864 9650 johnl@esegue.segue.boston.ma.us, {ima|lotus|spdcc}!esegue!johnl "Now, we are all jelly doughnuts."
fr@icdi10.UUCP (Fred Rump from home) (01/15/90)
In article <1990Jan5.174006.2262@esegue.segue.boston.ma.us> johnl@esegue.segue.boston.ma.us (John R. Levine) writes: >toma@attctc.Dallas.TX.US (Tom Armistead) writes: >>I have a "requirement" to provide hard facts regarding COBOL performance >>under UNIX. ... > >It would be interesting to see some Cobol performance figures for a 386. While I don't have any real numbers either, the one big COBOL package out there in Xenix/Unix land is Real World and MCBA. Both are decendants of the Mini world of old and run one hell of a lot of systems. We write our own tight C code for a vertical, but also recommend and install Real World into lots of systems. Speed does not seem to be a real issue in today's 25Mhz and up 386 boxes. But SIZE ... wow, do these COBOL programs chew up disk space. Lots and lots of files. It seems that programs that don't really do a lot still have all the overhead of COBOL again and again and again. fred -- This is my house. My castle will get started right after I finish with news. 26 Warren St. uucp: ...{bpa dsinc uunet}!cdin-1!icdi10!fr Beverly, NJ 08010 domain: fred@cdin-1.uu.net or icdi10!fr@cdin-1.uu.net 609-386-6846 "Freude... Alle Menschen werden Brueder..." - Schiller