Chat Private

[jazzz]

This PR adds a specification which describes a conversation protocol for pairwise communication called PRIVATE1.

It covers payload generation and operation after a key has been agreed upon by both parties.

Notes

This may eventually be simplified and use Reliable Channels but its explicit in how it utilizes SDS + segmentation and the reasons for doing so.

[jazzz]

Unsure exactly what the best course of action is here. Recent changes to SDS requires an sender_id be attached the the SdsMessage.

This then brings in a concept of identity to this specification. Its definitely possible to introduce Identity concepts in order to pass them through to SDS, however it seems like a lot of added complexity

[jonesmarvin8]

All of these github references from vacp2p and waku-org are broken.

[osmaczko]

Looks good overall. The reasoning behind the Content -> Segmentation -> Reliability -> Encryption layering makes sense.

Please remember to address the TODOs before merging.

[osmaczko]

Suggested change

	```mermaid
	flowchart LR
	Content:::plain--> PrivateV1 --> Payload:::plain
	classDef plain fill:none,stroke:transparent;
	```
	```mermaid
	flowchart LR
	Content:::plain --> PrivateV1
	PrivateV1 --> F1[Frame 1] --> P1[Payload 1]:::plain
	PrivateV1 -.-> F2[Frame 2] -.-> P2[Payload 2]:::plain
	PrivateV1 -.-> FN[Frame N] -.-> PN[Payload N]:::plain
	classDef plain fill:none,stroke:transparent;

[jazzz]

I generally like these changes, but looking to understand more to see if there is more optimizations to be made.

I see two changes here.

1. An explicit reference to Frames
2. An specific update the the cardinality of messages.

I can see # 2 being an improvement as it explicitly refers to segmentation being a responsibility of the protocol.

At a high level does adding Frames increase understanding at a 40,000ft view?

[osmaczko]

At a high level does adding Frames increase understanding at a 40,000ft view?

Don’t have a strong opinion here. Including frames makes it clear at first glance that segmentation happens in the chat protocol rather than in the delivery service.

[osmaczko]

This contradicts segmentation specs:

Messages whose payload size is ≤ segmentSize are sent unmodified.

[jazzz]

Great point.

I specifically have suggested a change in that PR.

Sending the message unmodified violates the functional requirement of "deterministic decoding".

To decode those messages one would first attempt to decode as a Segmentation message, and then on failure, assume its a different type.

I'd prefer to spend the extra 4 bytes and make it explicit.

[osmaczko]

I'd prefer to spend the extra 4 bytes and make it explicit.

Totally agree 👍