Reposts with deletions/edits V2

[moodyjon]

Context: https://github.com/lbryio/types/issues/49

This based on a different claim.proto schema that limits modifications to only reposts of streams and only within a specific set of "Stream.Extension" fields. This is enforced at the level of the protobuf.

The Stream.Extension message contains a protobuf.Any. Right now I check the type name to make sure it's a type that 1) is known when the SDK is built, and 2) comes from a limited list of messages that are intended to be used as extensions. But I think it would be possible to put a claim ID as the "type_url_prefix" of protobuf.Any and download the "descriptor" needed to understand the type. For example, if the "@type" field of the Any message contains "/pb.Cad", the SDK can try to download the contents pointed to by .

To see the claim.proto changes go here: https://github.com/moodyjon/types/commits/repost_modify

Hub changes that would index the extension fields are TODO.

[lyoshenka]

leaving a note since im assigned as reviewer: we decided to go with a simple string->[]string mapping instead of protobuf.Any

[moodyjon]

State of this is I need to remove the dynamic extension type stuff, and add some code to catch StreamExtensionTypeUnresolved when an unknown extension is encountered. For example, a stream claim has some extensions only some of which are of known type. We would want to ignore the unknown ones.

[moodyjon]

To see the claim.proto changes go here: https://github.com/moodyjon/types/commits/repost_modify

The latest is I am storing extensions in a map instead of a repeated field. (Got that idea from: https://github.com/lbryio/lbry-sdk/issues/2983)

    // Value should contain one of the extension types (e.g. StringMap).
    // Key provides a hint indicating what keys we should expect in the map.
    map<string, google.protobuf.Any> extensions = 14;

For these StringMap extensions, the key of the map entry is what I was previously placing in StringMap.schema. (For resource-like things, the key would be the resource name).

Also, the values in StringMap are lists of strings. (Got that idea from: https://github.com/bb-msf-dd/types/pull/1/files#diff-4d1a25cbe9fc3b724c4dd38caf452672c542838cdc0cd567dc12460804534fc5R97)

// StringMap is a generic kind of Stream.Extension (see claim.proto) which
// is useful for testing, but is not space-efficient and does not enforce a
// limited set of fields.
message StringMap {
    map<string, StringList> s = 2;
}

message StringList {
    repeated string vs = 1;
}

[moodyjon]

Latest .proto changes look like:

message Stream {
    Source source = 1;
    string author = 2;
    string license = 3;
    string license_url = 4;
    int64 release_time = 5; // seconds since UNIX epoch
    Fee fee = 6;
    oneof type {
        Image image = 10;
        Video video = 11;
        Audio audio = 12;
        Software software = 13;
    }

    // Map key is a label for the extension hinting what info we should expect.
    // For example, one could have 2 extensions, "ebook" and "literature", which
    // are both StringMaps, but contain different sets of keys.
    // Map value should contain one of the extension types (e.g. StringMap)
    // encoded as a protobuf.Any.
    map<string, google.protobuf.Any> extensions = 14;

    message Modifiable {
        // Limit modifications to extensions for now.
        map<string, google.protobuf.Any> extensions = 1;
    }
}

message ModifyStream {
    Stream.Modifiable deletions = 1;
    Stream.Modifiable edits = 2;
}

message ModifyingClaimReference {
    bytes claim_hash = 1;
    oneof type {
        ModifyStream stream = 2;
        // Limit modifications to streams for now.
    }
}

// StringMap is a generic kind of Stream.Extension (see claim.proto) which
// is useful for testing & prototyping, but is not space-efficient and does
// not enforce a fixed set of fields.
message StringMap {
    message Value {
        repeated string vs = 1;
    }
    map<string, Value> s = 1;
}

[moodyjon]

Planning some further simplification after talking to @lyoshenka.

[moodyjon]

Closed in favor of #3706