canonicalize protobufs somehow.

It would be nice if we could canonicalize typed objects somehow, so that logically identical values produce byte-for-byte identical blobs, allowing them to be correctly de-duped. I don't know if there's a standard way of canonicalizing protobufs? I wasn't able to find anything after a bit of searching.

Relatedly: @bobg, how wedded are you to protobufs in particular? If there is indeed no standard canonicalization algorithm, it may be worth considering other formats (I'm somewhat partial to capnproto).

The issue names no files, tests, or entry points. Start by locating the typed-object serialization and de-duplication paths, then investigate whether protobuf or another format can provide canonical bytes. Done means logically identical values consistently produce byte-for-byte identical blobs and can be de-duplicated.

Written by the indexing model from the issue text.