[Portland] Project Portland, RuDI and the Generic desktop adapter

[Martin Konold]

Am Freitag, 9. Dezember 2005 13:31 schrieb nf2:

Hi,

> A) Is it about breaking the low level "interface standard" of shared
> library interfaces (regardless of APIs).

This typically happens due to carelessness or more often due to a change in 
compilers (e.g. C++ ABI was not stable in the past with g++)

> B) Or does it also cover binary compatibility problems caused by
> changing APIs (function and struct signatures).

> I always thought it was A, because B can be avoided through API
> discipline.

No, this is not the full picture. From time to time you want to be able to 
change the API in an ABI incompatible way. For KDE we refrained from doing so 
for the entire KDE 3.x line. But to KDE 4 we intentionally want to break both 
the API and ABI in order to allow us to innovate.

In general my conclusion is that we cannot provide ABI stability for the 
desktop for something like 10 years.

Therefore I consider it important that ABI changes don't affect perfectly 
running application on one hand and on the other I want applications 
independent of the chosen implementation laguage (C/C++,Java,C#) run on any 
future version of the Linux Desktop.

Regards,
-- martin

-- 
http://www.erfrakon.com/
Erlewein, Frank, Konold & Partner - Beratende Ingenieure und Physiker