🚀 go-pugleaf

Richard RUDEK (POQ@NoMoreSpam.com) wrote:
: jan@orchide.zzdialix.com.au (Jan Pompe) wrote:
: >Richard RUDEK (NoMoreSpam@POQ.com) wrote:

: [snip]

: >: I'm sure there is a better way, but it will require completely new
: >: tools and attitudes. I don't think you'll see any revolution until
: >: after the new millennium. Maybe.
: >
: >No new tools aren't needed the ones we have will do.
: >As for new attitudes I think old ones revised and revived will see
: >some improvement.
: >
: >I won't hold my breath.
: >

: I talked about new tools because currently, C++ and/or the application
: frameworks produce way too much operational/runtime data that attach
: and hide themselves to the runtime objects.

The problem is not limited to C++: comming from an embeded/assembler
background I made a foray into M2 some years ago and found that it 
compiled many routines from the library modules, which weren't used
and I had to delete them from the code to fit the program into the
available rom space.

While this speeds up application development there is a memory
usage penalty. Application developers have found it an acceptable
trade off and that is why it's happening. Don't forget that those
who make the tools are software developers too, and subject to the
same sort off pressures. If there was to be and industry wide change
in attitude this would all change 
 
: And I don't think going back to non-object programming makes much
: commerical sense.

It depends entirely on the type applications being developed some
stuff will run better if procedure oriented methods are used.

: Pre-compiled headers help compilation speed, but there is just way too
: much overhead in there, and apparently nothing to distill it !.

see above

: Take a typical problem: predefined parameter constants. If you could
: define these constants as enumerations for a particalar parameter
: within a particaluar object, then an object browser/editor can easily
: determine the range of values a parameter can be for that object from
: live data. Not hardcoded as it is currently.

: Kind of like VB5 tries to do, but better and without the rats nest
: reinterative objects.

: Many frameworks I've seen resort to global/module constants: 

:   #define WNDMSG_OK   101

: or somthing and the like. Others create an object for practically
: every Parameter.

While not very familiar with OOP or C++ the above strikes me as 
plain sloppy. You've seen the problem though, and have a possible
solution why not turn your idea into a working piece of software?

: Likewise, I would also expect it to produce efficient Assembly code
: when compiled. I mean, that to me is what a high level language is
: suppose to do...

High level languages produce binary imaages efficiently if the
machine code is also efficient this is a bonus.

To give and idea of what can be done a friend of mine got hold
of the source for a spice program, written in Fortran, the binary
image to run on a VAX was 4M, he cleaned it up added some extra
features and recompiled it and it's running happily on his 
TRS 80.  (for the youngsters a 64k machine:-).
   
-- 
Jan

Before replying by e-mail consider how "dialix" is spelt.