exploration: Agoric marshal types, protobuf, rholang data types

[dckc]

As @erights was presenting Agoric marshal types and such in the ZEstival last week, I said it's remarkable how much it looks like the data types in rholang: immutable lists, sets, maps, etc. as well as matching and sorting.

Based on some some type-directed prototyping (marshal-rhoproto.js in service of rchain-community/js2rho#16 , I think using RhoTypes.proto for interchange is promising, though I doubt the matching and sorting semantics are consistent between Agoric and RChain in the details. I haven't tried a serializing matchers in such a way that they would show up as matchers in rholang yet. That would be a good test.

cc @cwebber @kriskowal

[zenhack]

Hm, there's a lot of stuff in RhoTypes.proto that probably don't make sense in language-agnostic protocol; it looks like it is designed to encode a full rho AST, not just data literals & caps.

...either way, we really should nail down the data model before we devote too may cycles to specific encodings.

[dckc]

Hm, there's a lot of stuff in RhoTypes.proto that probably don't make sense in language-agnostic protocol; it looks like it is designed to encode a full rho AST, not just data literals & caps.

sure; I don't propose to use the whole thing. Just the relevant parts.

...either way, we really should nail down the data model before we devote too may cycles to specific encodings.

I don't know how to nail down abstract things; only running code that I can test.

After all, .proto is abstract in that it has at least 3 concrete syntaxes. (JSON, text, binary)

[dckc]

doubt this is worth tracking going forward