Re: Is Ada likely to survive ?

[dewar@merv.cs.nyu.edu (Robert Dewar)]

Seymour said

  While there were some flaws, ALGOL 68 was a fairly nice language. IMHO
  what killed it was a defining document that was unnecessarilly terse.
  The use of a W-grammar to define the language was reasonable; the
  paucity of elucidation was not.

Robert replies

  I don't think unnnecessary terseness was the issue here, but rather
  the sacrifice of informal readability to precision and accuracy. There
  were VERY rarely any arguments about what the defining document said
  or meant, and very very few errors were found (though of course you
  could argue that it was never widely used compared to say Ada).

  But the definining document is no less accessible than say the COBOL
  standard in my view, and had the advantage that implementors at least
  could exactly understand what it said, and easily find anything they
  wanted there. Yes, it required some effort to understand, but that
  is true of any formal definition (by comparison, the Ada standard is
  an informal definition at best).

  The problem is that formal definitions are indeed very much inaccessible
  to those without the skills in reading documents of the type. No amount
  of informal explanation would have helped people to read the Algol-68
  RR if they did not have the necessary mathematical background to be
  comfortable with a highly formal style of presentation.

  For most languages, programmers never see the defining document. How many
  COBOL programmers have read the COBOL standard? Virtually none. How many
  C programmes have read the C standard? Very few. How many C++ programmers
  have read the standard? Trick question -- none, there is no standard yet.

  People typically learn languages from sources other than the defining
  document, and this has been true ever since Algol-60 days (Algol-60
  was the one clear counter-example). Ada-83 came close to Algol-60
  in the extent to which programmers had at least looked at the standard,
  but still most programmers did not learn Ada 83 from the RM.

  What went wrong in retrospect was that informal materials were far too
  slow in coming, and that compilers were slow to appear. If the Algol-68R
  compiler had received wide dissemination on many machines, together with
  Ian Currie's wonderful 80 page "little yellow book", the picture would
  have been different. Indeed in the UK, Algol-68 was widely used, and a
  survey in the early 70's in the British Computer Journal showed the
  overwhelming majority of British Universities naming Algol-68 as the
  ideal teaching language (that same survey shows US universities choosing
  Fortran as the ideal teaching language!)

  Unfortunately the ICL machines were never very successful outside UK
  (where the government forced many universities to choose them).

Seymour said

  SPITBOL is basically SNOBOL 4 with a few minor things stripped out; it
  is quite different from Icon. I understand that there is a SPITBOL for
  the PC from Catspaw.

Robert replies

  Not quite, they are dialects. Spitbol leaves out a couple of the most
  dynamic features of SNOBOL4 that are rarely used (e.g. the ability
  to distinguish between names and values dynamically in pathological
  cases), but also adds many useful features. Spitbol is a compiler 
  that runs typically ten times faster than the Snobol4 interpretors,
  and was one of the early demonstrations that highly dynamic languages
  could be compiled efficiently.