Hacker News Reader

I am confused. The major problem is claimed to be un/marshalling. But they take no closer looks. Advise an architecture without it just to introduce another with it.

Seems like protocol overhead will stay.

View on HN

Yeah, the only thing we can really learn from this paper is that the authors aren't very familiar with protobuf. Anybody who claims it takes 10 microseconds to decode a 1KiB protocol message is either intentionally constructing the worst-case message or doesn't know what they are talking about.

> Yeah, the only thing we can really learn from this paper is that the authors aren't very familiar with protobuf.

This is stupid. Three of the four authors are engineers at Google.

That should not surprise you. When Microsoft was at its dumbest, its corporate identity was that they were hiring the best of the best.

Now Google has that baton. Wonder who they’ll pass it to...

I did not intend to express agreement with the OP's stupid comment.

If marshalling is the problem, fix marshalling. Erlang (built in distribution), CapnProto, Flatbuffers, etc

https://capnproto.org/news/2014-06-17-capnproto-flatbuffers-...

Reply on news.ycombinator.com