Protocol recommendations to Thom for Artemis v3

[rjwut]

Thom is working on Artemis v3 with a new code base. I would expect that this would also result in a significant overhaul to the protocol. I think it would be good, while Thom is still in the early stages of development, to put a bug in his ear about things we'd like to see him do with the protocol. Let's use this issue to draft a list of requests to Thom with regards to protocol design. Here are some suggestions I have:

• Publicly release the protocol spec. Having official and complete documentation of the protocol would make it much easier for modders.

• Semantically version the protocol separately from the Artemis software. People consuming the protocol should be able to be confident that their code will work with any version of the protocol with the same major version number.

• Decouple enumeration values from their order, and don't rearrange or reuse values. Changing what any enumeration value represents makes things more complicated for people writing code that uses the protocol.

• Name packet types for what they do, not for their internal format. Future developments may require you to add fields or change their type, making the type name misleading; e.g. valueFourInts.

• If you significantly change the format of a packet type, deprecate it and create a new packet type instead.

Anything else?

[rjwut]

I've posted this to the Artemis forums.