The "question" was about incompatibilities caused by standard interpretations and Mr. Schmerler answer was about different Autosar versions - which shows more about (good) PR skills than about solution for real problems.
IMHO actual Autosar implementations/usage are at the same stage as early OLE/COM Windows applications - big mess. During design of software based on Autosar, the XML editor (for handcrafting configuration files) is one of the main tool and deep inside knowledge about Autosar intrinsic is just a must. This is something different than it was promised by Autosar.
After more than 5 years of intensive "standard" changes it is really hard to believe that Autosar have good vision what they (as a consortium) want to achieve.
Stefan Schmerler's comments reflect common thinking about standards, but not current theory.
It is possible to create standards for multi-layered programmable interfaces (e.g., APIs)that are perfectly backward compatible and do not "obstruct the possibilities for technological progress." A paper identifing how this is accomplished on interfaces for next generation networks is available at http://www.csrstds.com/pdf/exploring.pdf