Is there a place for 4GLs anymore?

[Jon (Jarod) Hawks]

I'll contrast for fun, class with my experience with a
vendor when I wrote at OS level.

1GL is machine language - 
binary 1011 1101 etc based on Hexadecimal, 
Some used Octal or Trioctal 377 377 001 a regrouping
of
 Hexadecimal; 

2GL is symbolic machine language such as assembler
    a variant on this was to get closer to a language
    useable for programming apps and special utilities
    typically meta-symbol or macro bal; virtually all
    languages broke down into this. All assembled to
    executable binaries - objects.

2.5GL BASIC - Interpreter languages, A compute engine
      interpreted and executed the language but the
      language was not really compiled to exec binary.
      PERL is an aberrant of the 1970's basic
statements
      with print macro's and now other programming
      interfaces built into its library.

3GL is high level procedural such as COBOL, C, C++,
Java
    Three GL for us was defined as a compiler
language;
    among them, Fortran, Cobol; Procedural code was 
    classed as a compile with prep, and required the
    execution to resolve at run time; variants SPL, 
    Pascal/Delphi, C, C++; SPL and Mod 8 have more 
    basis for Java language than Java, which is still
    evolving. The Binary Object Modules have taken
    different connotations but provide the
architecture
    for the Bean classes across the board. The DLL
    concept also spawns from here. Years ago, in the
    1990's Cognos made the trusted mistake of going
    into the labs with both Oracle and Microsoft, who
    took full advantage of Powerhouse architecture. In
    1993 the VP of Tech of Digital moved to Microsoft
    and also added Distributed COM. Java and
Powerhouse
    share a fair amount of architecture. HP has 
    provided a substantial amount of language
    architecture for these and other developing
    languages. Cognos was at least 20 years ahead of
    the pack.

4GL a non-procedural (what not how) language that
provides a large reduction in physical code. Most are
dictionary based. I have a contrary definition from
our
labs. 4GL was an attempted English-like language, then
non-procedural, to interface with users, much like SQL
was initially invented for. Transactional processing
came after the fact for both of these. Cognos managed
to make a logical structure and then, procedural in
the background. You don't see the structure because
it's
controlled by the background engine. Axiant also
inherits this characteristic, along with Impromptu. 

5GL was defined as AI. However, no language actually
emerged for this other than C, which was a defacto
from
Ux writers coming out of college.

We have seen a multitude of programmers that have
learned the Ux platform and languages but fail to
fully comprehend the necessary underpinning of the
database and how to associate tables with actual
processing. They don't understand the variants
involved with transactional and volume processing. The
sadest part of the diminished interest in 4GL is the
ETL functions exhibited by vendors like Informatica,
Sagent and others who have languages that are
primitive to the capabilities of Powerhouse. It would
be a great plus if Cognos provided HP with a standard
release of Powerhouse for their products. The small
cost upfront would be a source of continued revenue
for support and give Java competition on the midframe,
which is lacking, in large volume processing. Java is
very slow and the J2EE environments of WebLogic and
Websphere are not up to spec for fast processing of
large objects. I have to go to extensions, written in
CORBA, and was forced to use Visibroker, which is like
spreading your toilet refuse on your house lawn, but
it works. IONA, another canadian group, has a superior
CORBA ORB. Anyway, it would be great to see the
Powerhouse division, increase competition for the
mid-frame. As far as web programming, which has taken
root much like client-server in the 90's, it will take
some time for young folks to understand, that the core
of the logic, needs to be housed on the mid-frame. And
that the language needs to be less dependent upon so
many other changing parts and licensing. Companies
providing software, such as Borland, stink in the
repackaging, and force the buyer into new packaging. I
was approached by a salesman to buy 900K worth of
Borland licensing. I had to call the VP to get the
salesmen out of the way and came to a 90K solution.
All due to repackaging. The poor understanding among
IT managers being influenced comes from poor training
and experience in the programming realm. They're good
at HR but suck at technology, and push their way into
the contracting. Stupid, but now a reality. 

End of diatribe.

=====


__________________________________
Do you Yahoo!?
The New Yahoo! Search - Faster. Easier. Bingo.
http://search.yahoo.com