32

In his 1993 conference proceeding The Development of the C Language, Dennis Ritchie stated

Successors

C and even B have several direct descendants, though they do not rival Pascal in generating progeny.

Was Ritchie correct, or was he just being modest? As of the date of the quote (1993), were more computer languages descended from Pascal than C?

We are counting languages, not their specific compilers/implementations (e.g. UCSD Pascal, Borland C). Some languages may be descended from both.

(I doubt such a claim would still be true today.)

DrSheldon
  • 15,979
  • 5
  • 49
  • 113
  • 22
    It's difficult to say without knowing the context, but given Ritchie's opinion about Pascal, I'd interpret "they do not rival Pascal in generating progeny" as a tongue-in-cheek criticism along the lines of "Pascal was so bad it needed lots of successors to at least get a few things right". If this reading is correct, being "too modest" isn't what this is about... – dirkt Apr 25 '19 at 10:33
  • Aren't some features of C itself derived from Pascal? So if you're taking this definition in the loosest sense, all descendants of C are also descendants of Pascal, (if not direct descendants) so Pascal would obviously have to have more. – Darrel Hoffman Apr 25 '19 at 19:37
  • 2
    @Darrel Pascal and C are contemporaneous; features of C which look like Pascal are more likely to come from Algol (which both C and Pascal derive from). – Stephen Kitt Apr 26 '19 at 11:11
  • "and even B" -> does B actually have any known direct successors other than C? Nothing linked from this page lists anything else. I was under the impression that the distinction between B and C is itself largely retroactive and there wasn't really a clear demarcation at the time C evolved gradually from it, so much as that "new B" eventually needed a new name. – Alex Celeste Apr 27 '19 at 12:07
  • 2
    @Leushenko Ritchie mentions Eh and Zed as descendants of B. – Stephen Kitt Apr 29 '19 at 10:38
  • 1
    @StephenKitt thanks! I was able to find a reference manual for Eh here, after some amount of manual trawling; based on comments by the creators, it sounds like Zed is lost to history unless a hardcopy of the manual survives somewhere at UW. Apparently the machine hosting Thoth broke and was thrown away without being backed up! – Alex Celeste May 05 '19 at 14:11
  • Nice find @Leushenko! – Stephen Kitt May 05 '19 at 14:44
  • In addition to the many well known Pascal descendants listed, in the 80s many corporations had their own proprietary languages, which were sometimes based on Pascal. Two I can think of are BNR Pascal (BNR was the predecessor of Nortel) which they used for phone equipment. Control Data Corp had Cybil, which they used for all system software in the Cyber-180 line. – Theodore Norvell May 03 '21 at 19:04

3 Answers3

39

Was Ritchie correct, or was he just being modest?

I’m not sure modesty plays a part here; I don’t see any statement of value attached to the size of a language’s family.

As of the date of the quote (1993), were more computer languages descended from Pascal than C?

Yes, he was correct, at least as phrased in the paper. Perceptions are perhaps biased because people tend to consider Pascal’s family tree v. C’s direct descendants (as the paper suggests itself); the main direct descendants of C by 1993 are listed in the paper:

  • Concurrent C
  • Objective C
  • C*
  • C++

Expanding the search somewhat (see for example O’Reilly’s language poster and Éric Lévénez’ Computer Languages History), one could add other direct descendants which are arguably very similar to C, although I’m not sure Ritchie was considering them when writing the paper in question:

  • the C shell
  • AWK

By 1993, Pascal had a similar number of direct descendants (identical in this analysis, but my counts are no doubt partial):

  • CLU
  • Mesa
  • Modula
  • Ada
  • Object Pascal
  • Concurrent Pascal

but in my experience people tend to think of the Pascal “progeny” as including some indirect descendants, in particular the Modula and Oberon families (Modula, Modula-2, Modula-3, Oberon, Oberon-2), and variants of Pascal itself since “core” Pascal wasn’t really usable as-is so developers used platform-specific dialects.

As you mention yourself, later languages tried to integrate lessons learned from both C-related and Pascal-related languages, and they all derive from Algol anyway. (Pascal itself started off as suggested improvements to Algol X, which weren’t accepted into the language.) Languages where the family trees rejoin include Python and Icon.

The reason Pascal is at the root of a larger family tree hasn’t been evoked yet.

Pascal was designed by Niklaus Wirth, a researcher whose work focused on studying programming itself, including the use of programming languages as teaching tools and structuring tools (for structuring programmers’ minds); programming languages were nearly a side-effect of that research, and Wirth iterated on programming languages multiple times, which led to a profusion of languages and variants. Pascal was widely studied in academic circles, and others also pursued this type of research, leading to more descendants than Wirth himself could produce.

C on the other hand was designed by Dennis Ritchie, as a tool to develop Unix. It wasn’t a research vehicle, and thus didn’t immediately produce the wealth of variants and descendants that Pascal did. It initially spread with its users, not with papers describing it, at least until The C Programming Language was published (six years after the creation of C).

The programming languages genealogical tree project includes a nice diagram, and a large list of other resources on the topic.

Stephen Kitt
  • 121,835
  • 17
  • 505
  • 462
  • 12
    There's a nice quote about Wirth's intentions behind Pascal included on the Free Pascal Wiki: "Occasionally, it has been claimed that Pascal was designed as a language for teaching. Although this is correct, its use in teaching was not the only goal. In fact, I do not believe in using tools and formalisms in teaching that are inadequate for any practical task.". – TripeHound Apr 25 '19 at 07:56
  • 6
    I think it is much more accurate to say that Algol was the root of these languages, not Pascal. Pascal was largely created to be a better Algol after the debacle of Algol68, and there were a wide variety of other spin-offs that were in widespread use at the time - JOVIAL being an obvious example. – Maury Markowitz Apr 25 '19 at 10:33
  • 4
    @Maury basing the discussion on Algol makes it harder to create an artificial distinction between C and Pascal ;-). – Stephen Kitt Apr 25 '19 at 13:14
  • @MauryMarkowitz Since comments are probably not the appropriate medium, do you have any links that expand on what "the debacle of Algol68" was? (A quick Google search only threw up this answer on StackOverflow which only mentions the debacle but has no links (other than being somewhere in "the ACM History of Programming Languages conference papers"). – TripeHound Apr 25 '19 at 13:34
  • 3
    @TripeHound the Lindsey paper covers the debacle in detail. To give you an idea, it introduction starts with “The World seems to have a rather negative perception of ALGOL 68. The language has been said to be “too big”, to be defined by an “unreadable Report” produced by a committee which “broke up in disarray”, to have no implementations, and to have no users.” (The paper does try to redeem the language.) – Stephen Kitt Apr 25 '19 at 13:54
  • @StephenKitt Thanks for the link... I'll have a dig. From memory (I dabbled with Algol68 about 35 years ago, and quite liked it), the definition of the language was a fairly slim (¼"?) book and I don't remember it being unreadable... – TripeHound Apr 25 '19 at 13:58
  • 1
    @Tripehound - the problem with the Algol68 Report was that it used a new two-level grammar in a difficult way. The Revised Report redid the grammar in a cleaner way but it was still relatively complex for the time. Would the "slim book" you refer to be the Algol 68R User's Manual for ICL1900? That's very simple but is not the defining document. Slightly more precise (but still not the official definition) is Lindsey/van Der Meulen's Informal Introduction to Algol 68, which I rate extremely highly. – dave Apr 25 '19 at 14:19
  • @another-dave Yes, the "slim volume" was the manual for the ICL1900 (I was actually using a 290x but from what I remember it was mostly pretending to be a 190x). – TripeHound Apr 25 '19 at 14:26
  • 4
    Algol68 was widely used as the canonical example of second-system-effect in the 1970s. It was in all the language design books of the era. It added all sorts of huge distributed costs to implement features that no one actually used, yet left many of the original implementation problems from earlier versions unsolved. But all my refs are dead tree, but looking in Google Books you can find many comments as to its sheer size and complexity. – Maury Markowitz Apr 25 '19 at 16:35
  • 3
    Hoar's quote is best "There are two ways of constructing a software design: one way is to make it so simple that there are obviously no deficiencies and the other way is to make it so complicated that there are no obvious deficiencies." – Maury Markowitz Apr 25 '19 at 16:43
  • I appreciate that you've listed the descendants. However, you completely missed the various scripting languages which have descended from C. By 1993, this included C shell, ksh, tcsh, awk, and Perl. – DrSheldon Apr 27 '19 at 12:23
  • @DrSheldon indeed, I’ve tried to address that in my update. – Stephen Kitt Apr 29 '19 at 09:34
8

Short Answer: Yes

Was Ritchie correct, or was he just being modest? As of the date of the quote (1993), were more computer languages descended from Pascal than C?

(*1)

Well, it's not as easy as it sounds, as there is no birth certificate with all parents named. Both languages are Algol descendants. And throughout the 1980s Pascal was seen as the way to go - eventually all the way 'til Ada. So yes.

We are counting languages, not their specific compilers/implementations (e.g. UCSD Pascal, Borland C).

Err. No. This falls short of development. To start with, Pascal in its core definition is a very simple language (much more simple than C) intended for teaching classes about basic programming all the way to compiler building. As a result, it was so simplified that serious work was almost impossible - an elegant language for ivory tower games. It lacked strings, usable file I/O and any form of modularisation.

To make it usable for real world application, implementation had to enhance it substantially. And as usual, each and every developer had it his way. A UCSD Pascal program could not be compiled with Pascal/MT+ or Microsoft or Turbo Pascal, as all of them handled things differently.

If at all, then there is a main line promoted by the two most successful products: UCSD Pascal and Turbo Pascal, with the later being an extension of some sort to UCSD Pascal. The huge success of TP did lead to many other moving toward compatible constructs.

Heck, and then there are languages called Pascal which are derived from already different named languages, like Component Pascal evolved out of Oberon, which itself is a much improved Pascal child.

Bottom line, the distinction when it is a 'new' language or not cannot be made by the name.

Turbo Pascal and its incredible wide spread success (even I was tempted to use it) makes a good maker about how much more successful Pascal was in the 80s than C.

Some languages may be descended from both.

There is no 'pure' linage anywhere in the language world. To some degree all Algol based languages can as well be described as Pascal offsprings.

It may even be necessary to look past the syntactic sugar of brackets vs. keywords to realize that Pascal is mainly data driven, as its big step from Algol was the way to define data structures. In so far Ada makes the most 'pure' child of Pascal anyway (*2). Nowadays a feature to be found across basically all general purpose languages.

But let's try a list of close relatives openly carrying the family tradition:

  • Pascal
  • UCSD Pascal
  • Turbo Pascal
  • Pascal-SC
  • Oberon
  • Modula-2
  • Component-Pascal
  • Delphi
  • Concurrent Pascal
  • Pascal XL
  • Object Pascal
  • Pocket Studio
  • Vector Pascal (available for the PS2 !)
  • MS Pascal
  • Compaq Pascal (notable for type casting)

Besides all the variation two lines are notable. One is the Oberon/Modula development driven by Wirth, the other is the Borland Turbo/Delphi line. Each creating their own family of Pascal offspring sharing features. The later creating a linage looking like this:

  • (Pascal)
  • (UCSD Pascal)
  • Turbo Pascal
  • Borland Pascal
  • Object Pascal
  • Delphi

Well, and then there are all the languages with a less clear lineage all the way to JavaScript and Ada.

(I doubt such a claim would still be true today.)

As before, it depends on the way of counting. I'd say Pascal has still an advantage here :))

*1 - It may be noteworthy that Brian Kernighan in contrast offered quite some public (and less than correct) criticism about Pascal.

*2 - All the incompatible Pascal variations where the main reasons for the demand that Ada compilers could only be called that way if they adhere to a strict standard ... which eventually delayed Ada compilers way too long to really inherit the Pascal world, making room for C.

LаngLаngС
  • 1,596
  • 1
  • 9
  • 19
Raffzahn
  • 222,541
  • 22
  • 631
  • 918
  • 5
    IMO Turbo Pascal wasn't a commercial success because of Pascal itself, or the Borland extensions to it, but because of its IDE. – alephzero Apr 25 '19 at 10:46
  • Re Kernighan on Pascal: search for "Why Pascal is Not My Favorite Programming Language" and you'll find numerous copies. – dave Apr 25 '19 at 14:25
  • 6
    @alephzero "its IDE" sells it short. It was unbelievably fast. It allowed for frequent iteration in the code because of that speed. That was very much tied to the design of the language itself - namely, the fact that a single-pass compiler was trivial, and that you needed only a tiny amount of memory to store the whole thing. This also allowed for other huge features in the IDE - like the integrated debugger that just worked. Pascal had its shortcomings (most notably the lack of strings), but it was also very friendly, with a great community. People loved both Pascal and Turbo Pascal. – Luaan Apr 25 '19 at 15:16
  • 2
    @alephzero Back then, the low cost was also a factor. It could easily cost several hundred dollars for other development environments that weren't nearly as friendly. – MetalMikester Apr 26 '19 at 11:46
  • I dissagree that pascal is simpler, Pascal has more keywords and more operators than C and with can make things confusig. – Jasen Apr 27 '19 at 08:19
  • 1
    @Jasen It seams you are talking about a programmers PoV, which wasn't meant. It's about the compiler being more simple.Pascal follows a strict block concept with no exceptions, where C does restrict constructs to certain levels. For example in Pascal any block may contain a procedure, not possible in C, so the compiler needs to distinguish here, where Pascal just works ahead (Nested procedures are a nice way clarifying source code). Similar With can make sources way more readable than endless lines of nested structures needed in C. Ofc, this may depend on your style. – Raffzahn Apr 28 '19 at 20:38
  • 3
    @Luaan: More interesting than the integrated debuggger that was added in 5.0 was the way that 2.x and 3.x would identify the locations were runtime errors occurred. In both CP/M and DOS, programs started at address 0x0100, so if the user indicated that an error occurred at address 0x1234, the IDE would set the limit to 0x1134 bytes, disable compiler output, and run the compiler. The part of the program that was being scanned when the compiler hits its limit would be the location of the error. – supercat Apr 29 '19 at 15:38
  • 2
    @Jasen The number of keywords isn't a good measure of how simple a language is. Would you consider machine code simpler than C? It didn't have any keywords at all! :) I doubt you'd find many people who used both C and Pascal who'd consider C the simpler of the two. Indeed, it was a common complaint coming from C programmers that Pascal is too "simple" (usually followed by "so it encourages bad programmers"). The important thing is how easy it is to do the things you want to do. Of course, C was still closer to Pascal than C++ to Delphi in that regard, but that's out there :) – Luaan Apr 29 '19 at 17:30
  • I'm not sure Ada's standard compliance was the reason C took over; a world of Ada subsets was likely to have its own problems, and Ada was large for microcomputers that were taking over. I've also heard that Ada was generally sold and priced for military use, and never got the cheap compilers until 1995 and GNU Ada. – prosfilaes May 27 '20 at 08:28
  • @prosfilaes It was, as using the name Ada was tied to standard compliance, and there was no subset. So developing a basic Ada compiler was simply not legal. Or it could at least not be called Ada,undermining sales. At the same time Pascal (and Modula) compilers were standard for micros. Ada isn't much more complex, especially when starting with a core version and develop from there. Starting from scratch to a feature complete implementation is costly, nothing a small company can do - and the result will have a certain price tag only large customers could pay, shrinking the market even further. – Raffzahn May 27 '20 at 15:13
  • @Raffzahn Yes, it wasn't legal to produce a subset of Ada and call it Ada. I recall a review of an Ada-subset compiler that got around that somehow, and the review was upset about the implemented language barely larger than Pascal. I still feel Ada was a bit too large for the time and a bit clunky, and C/C++ had Unix and working code, and that removing that restriction wouldn't have changed anything. – prosfilaes May 29 '20 at 06:05
  • Note that c also lacked strings and usable file IO. Both pascal and c had character arrays, pascal had language-defined file IO, c required the operating system library to provide file IO. The argument was active at the time about if a language should include file IO (pascal) or defer it to the system library (c). After the design decision (internal vs external), discussion then devolves to the usefulness of the unix system library (later standardized as the c standard library), vs the usefulness of the inbuilt pascal file IO. – david Feb 20 '23 at 22:27
6

At the time he wrote that, it would have been literally true. Concurrent Pascal, *Pascal, SUE (which I'd like to know more about), Modula (not Modula-2), Modula-2, Mesa, Modula-3, Euclid, Concurrent Euclid, Turing, Turing+, the Gypsy specification language, LIS (another one I'd like to know more about), and arguably CLU. Those are just ones I can call to mind.

raok
  • 356
  • 1
  • 2