Contains a few corrections as well as revisions for upcoming protoc changes

[jhump]

This revises a couple of sections with extra information about protoc behavior:

1. For resolving relative references, extension names inside message literals are resolved a little differently. I'm pretty sure this is because the it uses the text format implementation for that, which is different from symbol resolution in the compiler. In summary: extension references in message literals do not support examining enclosing message scopes, only package scopes.
2. Messages may not set the message_set_wire_format option in proto3 syntax.

This also adds some content based on upcoming changes that will be in v22 of protoc (protocolbuffers/protobuf#11349 and protocolbuffers/protobuf#10586):

1. Custom JSON names are checked for conflicts. Also, the JSON name change is now case sensitive. (Previously, lower-cased names were compared.)
2. Reserved names are just string literals in the language, but now they are verified and must represent valid identifiers.

[vercel]

The latest updates on your projects. Learn more about Vercel for Git ↗︎

Name	Status	Preview	Updated
protobuf-com	✅ Ready (Inspect)	Visit Preview	Jan 21, 2023 at 2:58PM (UTC)

[jchadwick-buf]

Am I correct in understanding that this check in protoc is just a warning for now, or did it change since the PR?

[jhump]

You are correct, it is a warning in protoc. But it will be promoted to an error in the following release.

It was agreed that the existing behavior was incorrect, so I think it's okay for this spec to vary from actual protoc behavior if it is an acknowledged defect in protoc. In fact, even their spec already implies that this not being a valid identifier would be a syntax error. But they wanted to give users at least one version where it was a warning, to give an opportunity to see the warning and fix sources before this became a compile error.

FWIW, buf already treats this as an error (and we've had had no complaints from users on the discrepancy with protoc).

[jchadwick-buf]

I am curious about if the reserved names behavior change is really an error or if it's just a warning, but since it's not actually explicitly stated in this PR anyways I feel comfortable just giving the LGTM here.

[akshayjshah]

Love that we're getting ahead of future protoc releases! Nice work keeping up with G, Josh.

[timostamm]

Just a personal thank you for improving the JSON name validation! I always felt bad about pointing users to this feature because it was so easy to make a silly mistake and only notice it much too late.