Proposal: ensure amino - proto3 compatibility & developer convenience

[liamsi]

It looks like we are going through with #156 and #163 👍

I have a 2 thoughts I wanted to sketch down here even if it might be too early:

1. To actually and visibly ensure compatibility, it would be great to have script/triggered via CI that runs tests on proto files included in golang protobuf. These tests are also useful to document where amino still differs from proto3. Although I mentioned above that this might be too early, the earlier we have this, the better!

2. A feature for usability/to speed up development and prototyping: At EPFL/dedis we also had our own reflection based protobuf library. It was aiming for full compatibility and the main reason we had it was the convenience of just writing go code when prototyping (instead of having to write .proto files); that was great for rapid prototyping. It also had the possibility, to generate .proto files. I think, it would be awesome to provide similar functionality in amino. Then, devs coming from other languages and not familiar with amino, could just reuse our generated .proto files to write implementations in other languages. This could boost adoption and remove any barriers, too.

I'm happy to work on both issues here and later in amino_rs (in the latter we could even use the generated .proto files via 2. to generate rust code via prost! e.g. for kms, instead of coding this our self, just an idea)

Thoughts? @jaekwon @zmanian @ebuchman @mdyring

[ebuchman]

Wow cool. Do you think Dedis might be interested in collaborating on Amino ?

If we had proto3 compatibility, and could generate .proto files, we could just use Amino for ABCI :D

[liamsi]

Wow cool. Do you think Dedis might be interested in collaborating on Amino ?

Not sure, but I can ping @bford @ineiti. And definitely, that would be great! I think they are quite busy with writing awesome papers and receiving best paper awards :-D But last time I checked the team was developing another encoding approach called fixbuf. So, maybe amino might be an interesting alternative or addition to dedis/protobuf?

[jaekwon]

I think Amino can be a gateway to Amino2 (which is basically the Amino before this Proto3 refactor). There are merits to the original design, esp with regards to handling nil vs empty objects (Proto3 has no distinction), and more compact lists, and faster for most use-cases esp for blockchain processing.

Will try to read up on fixbuf.

Everything you wrote sounds great. I have a bit of logic written for AST surfing... should we start from that or?

[jaekwon]

I think they would be interested in go-amino because it's semi-mature and it achieves the goal of go-integration.

[jaekwon]

Just noticed that the dedis code is GPLv2. Any chance we can change that?

[liamsi]

Just noticed that the dedis code is GPLv2. Any chance we can change that?

Interesting, I though we switched everything to GPLv3 (as it allows patenting and is more friendly for commercial use). I can imagine that protobuf was just forgotten (e.g. fixbuf is GPLv3). (/cc @ineiti)