[Gllug] Linux on Desktop

Nix nix at esperi.org.uk
Wed Feb 7 00:14:45 UTC 2007


On 5 Feb 2007, Pete Ryland verbalised:
> The whole "third-party apps look crap and don't upgrade well" thing is
> an interesting thing to study carefully.  Which I did when I went from
> Win3.1 to Win95.  I had an app which worked well under Win3.1 that
> looked utter crap on Win95.  I used a debugger to inspect the API
> calls that it was making.  I then used the same technique to inspect
> the API calls that MS Word was using.  I found that MS Word didn't in
> general use any of the documented API, whereas the third-party app had
> used it throughout.  From one Windows version to the next, the
> undocumented parts of the API seemed to produce stable results, but
> the documented API calls changed behaviour in subtle but damaging
> ways, making the third-party apps look and feel much worse.  In
> summary, I think it's a conspiracy.

Never assume conspiracy when you can assume incompetence. If programmers
in MS are anything like the ones anywhere else, they'll ignore the docs
if they can grep the source tree, and will pay no attention at all to
whether a function is internal or not: if they need it, and they can
see it's there, they'll call it.

Then, when MS reaches its testing cycle, they test the hell out of their
own apps and test other apps rather less hard. Of course this means that
the APIs that own apps use will tend to get more tested than the
documented ones that nobody could be bothered to use.

(And then in the next release you get another layer of undocumented APIs
under the previously-undocumented ones --- and then Word et al start to
use those, and the cycle repeats, and before you know what's what you
have a stack 53 layers deep, and XP-scale bloat.)

-- 
`In the future, company names will be a 32-character hex string.'
  --- Bruce Schneier on the shortage of company names
-------------- next part --------------
-- 
Gllug mailing list  -  Gllug at gllug.org.uk
http://lists.gllug.org.uk/mailman/listinfo/gllug


More information about the GLLUG mailing list