Schemas for "internal data"

[Matt Stern]

Hello!

Suppose team A produces data with the following schema:

struct Data {

  id @1 : Uint64;

  stuff @2: InternalData;

}

Now, team B consumes this data but doesn't want care about the "stuff" field, and team B does not want to include the schema files for InternalData. So in the consumer code, this schema is used instead:

struct Data {

  id @1: Uint64;

  stuff @2: Void; // Reserved for team A internal data.

}

Is this safe to do? If it's relevant, team A produces with C++ and team B consumes with Java.

(For the record, InternalData is not privacy sensitive or anything like that. I know that team B could peek into that struct even without the schema and reverse engineer).

Thanks!

[Ian Denhardt]

Quoting Matt Stern (2021-02-16 14:21:43)

> Is this safe to do? If it's relevant, team A produces with C++ and
> team B consumes with Java.

No; the representations of Void and InternalData are different (I assume
the latter is a struct), so you'll end up with two types that are
wire-incompatible.

Rather than using `Void`, you can use `AnyPointer`, in which case this
will work provided that `InternalData` is some pointer type (struct,
list, or interface).

[Matt Stern]

Thanks Ian.

Yes, InternalData is a struct. I'll give AnyPointer a look!