Definition of a valid protobuf version for each OSI release

[TimFrickeBMW]

For each OSI release, one single protobuf version should be defined as "valid" or "to be used".

We have experienced that, without such a definition, incompatibilities arise between simulation tools and OSI consumers. Any bilateral agreements between OSI providers and OSI consumers on protobuf versions, however, would impede provider/consumer exchangeability.

(Note that this would not prevent anyone to use a different protobuf version, but it would set a standard.)

[pmai]

As far as I can see this is a problem that is specific to Linux (using OSMP), where running multiple versions of ProtoBuf in one process is problematic due to shared library usage. For e.g. Windows, where ProtoBuf can be statically linked, this should not be a problem - dito for Linux not using OSMP (e.g. models are part of seperate processes), right?

Or did you see incompatibilities of on-the-wire encodings - which would be a bug in ProtoBuf?

I'm asking, because then the recommendation can be tailored to that particular problem / use case.

[TimFrickeBMW]

We have not identified the root cause of our concrete problem. It could have been a conflict of library usage or a general inconsistency. Even though ProtoBuf is formally upwards and downwards compatible, using different versions on two sides of an interface can lead to unpredictable failure cases.

[TimFrickeBMW]

Even if not explicitly recommending a ProtoBuf version, every new OSI release could state "tested with ProtoBuf version X".

[PhRosenberger]

We face exactly the same issue in our lab. How will this be handled in the next release?