Re: database market share 2003

Noons wrote (in part):

> The whole single code base thing is quite bogus, IMHO. Just exactly
> when is a product single code base? I mean, in my time I've written a
> fair amount of #define and #ifdef pairs to try and make things work
> properly everywhere, and that was JUST in the UNIX environment. You
> should have seen some of my DOS/VSE Assembler macros! I wouldn't
> recommend that type of coding to anyone, but it had to be done to get
> over the problems way back then.
>

I used to have to write portable (among UNIX systems) code and as the number of systems went up, the number of #ifdef #endif pairs went up as well, to the point that code became unreadable (even though it worked). Since we used makefiles anyway, we changed things around so that the differences were mainly in header files instead of the #ifdef #endif pairs. And the makefiles ensured the correct header files were included.

The stuff I was working with was related to assembly-level optimizers, so that was not as easily done. In that case, the makefile determined from the target what source files to use as the code for an AT&T 32100 chipset differed too much from an Intel 80386 from a Motorola 68030 from a SPARC from a MIPS ...: you just could not use the same assembly level optimizer for each. But there was a lot of common stuff even so.

-- 
   .~.  Jean-David Beyer           Registered Linux User 85642.
   /V\                             Registered Machine   241939.
  /( )\ Shrewsbury, New Jersey     http://counter.li.org
  ^^-^^ 09:05:00 up 1 day, 18:30, 4 users, load average: 4.20, 4.12, 4.12