>>>>> "DE" == David Essex <[EMAIL PROTECTED]> writes:

  DE> Uri Guttman wrote:
  >> ...
  >> what about the runtime libraries for those cobols? i worked on PL/I
  >> libraries and they have many similar features to cobol (as pl/i was a
  >> genetic monster of cobol/algol/fortran). stuff such as isam record i/o,
  >> picture variables, decimal math, etc are needed for a full cobol. do
  >> those compiler provide that or are there libs provided for it? some of
  >> the fancier compilers of that generation actually generated inline code
  >> instead of many calls to runtime libs.

  DE> Yes, run-time libraries are included. They are written in C.

  DE> Generally, both OC and TC generate inline code for program flow
  DE> control. The RTL are used mostly for type conversions, system calls
  DE> and such.

then the question is how hard would it be to retarget from c to parrot?
the code generation will be very different (c is relatively easy to
generate especially for rtl calls). 

  DE> The COBOL standard requires 18 digits for declared and 31 digits for
  DE> intermediate variables (33 for COBOL 200x) of precision. On 32 bit
  DE> architectures this is generally emulated in software.

ooof.

  DE> For ISAM support both TC and OC use Berkeley's DB (now
  DE> Sleepyc*t). While functional, ISAM support has not been fully
  DE> tested. Nor does it meet the ISAM standard.

we wrote our own b-tree for isam as well as all the rest of the
voluminous rtl for pl1. c-isam would be a commercial possibility (are
they still around?). i thought sleepycat did hash indexes only and
btrees are really what is needed for isam.

  DE> Finally, while TC has a primitive transaction processor,
  DE> concurrency issues have not been addressed.

hmm, pl1 didn't have transactions IIRC. i forgot (never did learn real
cobol) that cobol did. that would be hard to support without major help
in the isam lib. if you really want to support real cobol in parrot to
handle all that legacy code, this will need to be addressed. but there
is a good commercial market for this (there is so much legacy cobol out
there and it can't be ported to other cpus) so maybe someone can make a
biz out of this.

uri

-- 
Uri Guttman  ------  [EMAIL PROTECTED]  -------- http://www.stemsystems.com
--Perl Consulting, Stem Development, Systems Architecture, Design and Coding-
Search or Offer Perl Jobs  ----------------------------  http://jobs.perl.org

Reply via email to