wapd@houxj.UUCP (Bill Dietrich) (11/08/84)
I have almost no experience with COBOL, but I thought it was very well standardized, compared to most languages. I thought there were 7 levels or packages, an ANSI standard, and so on. But I was reading the Nov 84 Unix/World, and the article "A Software Vendor's Search" by Neal Nelson says "... I had written applications in Data General COBOL, IBM Series/1 COBOL and IBM 370 COBOL. The manufacturer and operating system had made these three versions of COBOL so different that moving from one machine to another was almost as difficult as starting over in a new language." Is this typical of different COBOL versions ? Bill Dietrich houxj!wapd
johnl@godot.UUCP (11/11/84)
My impression is that Cobol is the best standardized of widely used languages. Most Cobols have lots of implementation-specific extensions to deal with screen packages, telecommunication monitors, and such, and I suspect that the guy who was complaining wrote his programs without regard to what was standard and what was not. I've heard hearsay of Cobol programs many thousands of lines long being ported changing only the environment division (a dozen lines, perhaps) which is defined to be implementation specific. John Levine, ima!johnl PS: How many other languages let you write: IF KICK-BACK GREATER THAN 10 AND LESS THAN 100 THEN ... without having to mention kick-back twice? They don't design them like they used to. Probably just as well.
henry@utzoo.UUCP (Henry Spencer) (11/13/84)
> My impression is that Cobol is the best standardized of widely used > languages. Certainly it is the *most* standardized, in a specific sense of the word: as Bill Plauger pointed out in Washington, since the COBOL standard has something like 12 different pieces that can be present or not in any given implementation, there are 4096 different standard COBOLs. That's easily the most standards of any language I know. Poor old FORTRAN only has two (well, three if you count FORTRAN 66). -- Henry Spencer @ U of Toronto Zoology {allegra,ihnp4,linus,decvax}!utzoo!henry
rcd@opus.UUCP (Dick Dunn) (11/13/84)
> I have almost no experience with COBOL, but I thought > it was very well standardized, compared to most languages. > ... > But I was reading the Nov 84 Unix/World, and the > article "A Software Vendor's Search" by Neal Nelson > says "... I had written applications in Data General COBOL, > IBM Series/1 COBOL and IBM 370 COBOL. The manufacturer > and operating system had made these three versions of COBOL > so different that moving from one machine to another was > almost as difficult as starting over in a new language." > > Is this typical of different COBOL versions ? COBOL is certainly standardized, but there are two problems: It has a real peculiarity--implementors are allowed to add reserved words! Given the structure of COBOL, most extensions will require new words. Given the application area for COBOL, there's a reasonable chance that new reserved words will collide with the identifiers used in a program written for a different implementation. It's an old language. The older a language gets, the more extensions are added, to solve new problems or solve old problems better. The extensions differ, so you have to be careful about what "features" you use if you're trying to write programs which are both portable and reasonably efficient. The second problem can be battled with care and a standard-checking compiler. The only hope for the first one (something I attacked once) is to collect a huge list of all of the reserved words on all the implementations you can find out about, and avoid ALL of the words. (The list will be in the hundreds of words.) -- Dick Dunn {hao,ucbvax,allegra}!nbires!rcd (303)444-5710 x3086 ...Never attribute to malice what can be adequately explained by stupidity.
dre@ptsfa.UUCP (Doug East) (11/14/84)
> PS: How many other languages let you write: > > IF KICK-BACK GREATER THAN 10 AND LESS THAN 100 THEN ... If I'm not mistake, which I could very well be, this particular nicety is not ANSI standard but is implemented on many compilers. It was on the CDC Cyber 7xx compilers along with some other very nonstandard stuff like reference modification (ala FORTRAN-like): MOVE THIS-IS-A-STRING ( 5 : 8 ) TO THIS-IS-A-SUBSTRING. And a very nice END-IF statement for the IF-THEN-ELSE structure which helps one keep track on the #?*% period. It also closed off a nested IF so that it would behave properly without the funky NEXT SENTENCE stuff: IF WHO-KNOWS-WHAT > ALOT THEN IF ALOT NOT > A-LITTLE THEN DISPLAY "ALOT IS TOO LITTLE" END-IF ELSE DISPLAY "WHO-KNOWS-WHAT KNOWS LITTLE" END-IF. Rather than: IF WHO-KNOWS-WHAT > ALOT THEN IF ALOT NOT > A-LITTLE THEN DISPLAY "ALOT IS TOO LITTLE" ELSE NEXT SENTENCE ELSE DISPLAY "WHO-KNOWS-WHAT KNOWS LITTLE" END-IF. However, all in all, I'm glad I don't program in it any more. Doug East (Pacific*Bell -- San Francisco) {ihnp4,ucbvax,cbosgd,decwrl,amd70,fortune,zehntel}!dual!ptsfa!dre
rdf@latour.DEC (11/14/84)
COBOL is indeed the most standardized (?) language I have seen, but in the case of companies like IBM, DG,DEC and others, many extensions have been made which do inhibit transporting software. Make no mistake though, DG's "screen section" for on-line COBOL programming, DEC's VAX COBOL extensions , and IBM's little "ON 1" all help sell more hardware for each company. Its up to the customer to not use extensions and stick to the standard when developing applications that they even think would have to be ported to a different O/S someday. As was said before, all you sometimes have to change is the ENVIRONMENT DIVISION, but this would be a well thought out applicaton. Standards are great, but if extensions are allowed, then compiler writers should not be the only ones looking at them. Rick
eugene@ames.UUCP (Eugene Miya) (11/19/84)
[]
Interesting following this discussion.
I, too, have heard that COBOL is the most standardized language, and I
have spent some time sitting on sitting on another ANSI committee, but
my experience with the COBOL language tells me that COBOL standardization
is something of a joke.
Several years ago, a large research lab was analysing the conversion of
software to newer systems. This facility had a large Univac and a slightly
smaller IBM population. The Univac people were in the seats of power, but
the important administrative work was done on a 3032. As a research
project using Maurice Halstead's Software Science measures, we ran
millions of lines of Univac FORTRAN and IBM COBOL thru counting
programs to 'analyze' the nature of shifting software from one manufacturer
to another. I found that IBM COBOL was "closer" to COBOL standard
with about 300 reserved words than Univac COBOL with about 500 reserved
words.
In the end we had to dump the analysis of the IBM->Univac conversion
because the cost of adding the 200 additional keywords was beyond
research budget (we had all that FORTRAN to analyse). These programs
did not just count line of code mind you, and we knew that languages
were more than the sum of the reserved words, but I thought I
would add some quantitative evidence to the COBOL discussion.
[In the end, the Univac forces `won' but the COBOL still sits on an IBM.]
--eugene miya
NASA Ames Res. Ctr.
{hplabs,ihnp4,dual,hao,vortex}!ames!aurora!eugene
emiya@ames-vmsb.ARPAmwm@ea.UUCP (11/20/84)
/***** ea:net.lang / ima!johnl / 8:22 pm Nov 17, 1984 */ PS: How many other languages let you write: IF KICK-BACK GREATER THAN 10 AND LESS THAN 100 THEN ... without having to mention kick-back twice? They don't design them like they used to. Probably just as well. /* ---------- */ At least three. The (much more readable, to my mathematically warped mind) line: 10 < KICK-BACK < 100 will work correctly in both BCPL and Icon. I'm not sure how BCPL does it, but Icon has the notion of "success" and "failure" of an expression. For instance, the statement if x < 10 then a :=: b will exchange a and b iff the expression "x < 10" succeeds. This happens if x is less than 10, and the expression has the value 10. Hence, 10 < x < 100 is either 100, or fails. In LISP, the expression (< 10 KICK-BACK 100) should work. <mike
cjl@iuvax.UUCP (11/27/84)
>it, but Icon has the notion of "success" and "failure" of an expression. >For instance, the statement >if x < 10 then a :=: b >will exchange a and b iff the expression "x < 10" succeeds. This happens >if x is less than 10, and the expression has the value 10. Hence, >10 < x < 100 is either 100, or fails. The relational operators in ICON return numbers, not boolean values. Actually there are no explicit boolean values in ICON but faillure, and success. Therefore 3 < 8 > 1 or (3 < 8) > 1 are evaluated successfully, but 3 < (8 > 1) return failure !!!! The reason is 8 > 1 returns 1 not 8. I cannot agree ICON's approach is easy to understand because this kind of language semantics is not natural and used in nowhere but some programming languages where there is no explicit boolean values. C.J.Lo ARPA : cjl@Indiana@CSNet-Relay UUCP : ...!iuvax!cjl