Joe Tsai

The following is a list of declarations that appear in the binary for a blank import of "encoding/binary": ``` reflect.resolveNameOff reflect.resolveTypeOff reflect.name.name reflect.name.readVarint reflect.name.data reflect.add reflect.Kind.String reflect.(*rtype).uncommon reflect.(*rtype).Kind reflect.(*rtype).String reflect.(*rtype).nameOff...

@beoran: Perhaps. One possibility is to put such functionality in the `math/bits` package: ```go func LoadUint32LE(v [4]byte) uint32 func StoreUint32LE(v uint32) [4]byte ``` See https://github.com/golang/go/issues/42958#issuecomment-751428308

After https://go.dev/cl/419757, a blank import of "reflect" bloats a binary primarily in two ways: * The transitive dependency on "errors" links in "reflectlite" (see #54341) * It has a dependency...

To work around this, we introduced a hacky API to specify the command name in `logpolicy` in https://github.com/tailscale/tailscale/pull/5625.

I'd love to see us maintain a single connection to some tailscale backend server and multiplex all of the logical connections on top of that, whether it be the control...

It's not so much a "community-written spec" that I see as an issue, but that there needs to be a formal statement that the spec is **the authority on correctness**,...

Related problem: ``` message M { optional string foo = 1; reserved "foo" "bar"; } ``` Again, the intention is to specify `"foo"` and `"bar"` as two separate identifiers. However,...

Also, the current behavior goes against the grammar specification, which says that the token should be a `fieldName` which is equivalent to the `ident` token. Related: https://developers.google.com/protocol-buffers/docs/reference/proto3-spec#reserved

This is working as intended. See https://github.com/protocolbuffers/protobuf-go/releases/tag/v1.20.0#v1.20-nil-values for details.

I'm not sure I understand why it's flushing every 240 bytes. The compressor already buffers a relatively large window before flushing since there's already a built-in buffer of at least...