compatibility vs. correctness

[Dmitry Gorbachev]

I believe that people wish "perfect world" API as an optional feature, not 
that they really want to broke compatibility.

[Steven Edwards]

On Mon, Dec 8, 2008 at 7:52 AM, Dmitry Gorbachev
[e-mail] wrote:
> I believe that people wish "perfect world" API as an optional feature, not
> that they really want to broke compatibility.

We discussed something like this years ago. It would be nice if every
sort of legacy API that has any sort of hacks for brokeness as part of
its implementation, could be isolated and a simple recompile with
something like a #define NO_LEGACY_WIN32 so you could recompile
without the hacks. Not really that useful for the real world as you
need the backward compatibility but it is a nice idea for embedded or
non-normal windows targets like ARM.
-- 
Steven Edwards

"There is one thing stronger than all the armies in the world, and
that is an idea whose time has come." - Victor Hugo

[Sarocet]

Steven Edwards wrote:
> On Mon, Dec 8, 2008 at 7:52 AM, Dmitry Gorbachev
> [e-mail] wrote:
> > I believe that people wish "perfect world" API as an optional feature, not
> > that they really want to broke compatibility.
>
> We discussed something like this years ago. It would be nice if every
> sort of legacy API that has any sort of hacks for brokeness as part of
> its implementation, could be isolated and a simple recompile with
> something like a #define NO_LEGACY_WIN32 so you could recompile
> without the hacks. Not really that useful for the real world as you
> need the backward compatibility..

Useful for the developer world. If I know that 'running in ReactOS'
means that the program is
Win32 strict, I'll favour developing on ROS, as then I will make better
code.
Another option would be instead of just failing, sending out a warning
on stderr whenever a
compatibility hack is applied.