Adrian Cole

I've disabled the vertx tests as it is a much bigger mess trying to get modern gpg2 working properly on an OS dist that went EOL over a year ago...

we redid our dependency management to ensure there's no conflict we cause wrt versions that should be managed by vert.x and sadly it still crashes https://github.com/openzipkin/brave/pull/1275

make it popular and sounds good!

the primary concern of reactor is propagation of context. there's currently some code hiding in plain sight in sleuth around that, originally a deviation of how we handle rxjava here....

just to know about existing practice. on the java side we have "parsing callbacks" which allow people to add whatever. doing more generically can prevent whack-a-mole changes later, as we...

also another option. if this is static.. with a span finished handler you can add another tag. ex as long as grpc client adds grpc.method tag you could look for...

wondering if this is a blocking change to move off our fork? considering all libraries discussed are not api stable (even http is 0.2.4) @lizan, there's an implication that there...

here's a suggested way out. 1. identify the data that caused the panic, assuming it was customer related and not a bytes generator 2. add a test that proves this...

PS on the point of the actual change here, I would be really cautious about any binary api especially looking at the broader picture and impact of change. I've had...

last note mainly to brain dump.. If it is really important non-interference, I would classify that differently than an api. For example if there are headers in as bytes and...