[
https://issues.apache.org/jira/browse/HTRACE-164?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14533211#comment-14533211
]
Colin Patrick McCabe commented on HTRACE-164:
---------------------------------------------
bq. Take this example: You ship this code. Then we want to add the ability to
ship incomplete spans ( Something that really happens in real world tracing )
So end ms needs to be optional. start ms need to be optional too The code
creating the bytes to write on the wire now needs to be changed by hand.
Different and de-coupled code on the server now needs to be changed to
understand that end_ms is optional
{{start_ms}} and {{end_ms}} already are optional in the current patch. msgpack
sends a map. If the 'b' key is not present, then {{begin_ms}} defaults to 0.
Similarly with 'e'. The same is true for all the span fields. They are all
optional at the RPC level.
bq. Or let say we want to add the ability to respond with if a span is durably
written somewhere. Or we want to respond with result saying to sample fewer
traces. Or we want to have the client able to set a trace as interesting after
it's completed.
If we want to add any of those things, we can add this to the
{{WriteSpansResp}} message in a backwards-compatible fashion. It's no
different than adding optional fields to a protobuf response message.
bq. Bug fixes are the same. There have been several times that msgpack has
gotten unicode incorrect. If we needed to change the type of a field that would
be a one off thing. With protobuf and thrift there are set rules around how to
do that in compatible ways.
It's not really that "msgpack got unicode incorrect", it's just that they left
the choice of string encoding up to the programmers, which was a bad idea. But
if you just use common sense and make everything UTF-8 there are no problems.
bq. There seems to be official C binding: https://github.com/msgpack/msgpack-c.
Any reason to prefer CMP to msgpack-c?
CMP is two source code files, which we can easily embed into our code.
msgpack-c is many more, and would require an external dependency or 3rd-party
build process.
[~abec] and I are giving an HTrace talk today from 2:20pm-2:40pm at HBaseCon.
Are you guys available to meet after the talk (i.e. at 2:40pm) to talk about
this?
> htrace hrpc: use msgpack for serialization
> ------------------------------------------
>
> Key: HTRACE-164
> URL: https://issues.apache.org/jira/browse/HTRACE-164
> Project: HTrace
> Issue Type: Bug
> Affects Versions: 3.2.0
> Reporter: Colin Patrick McCabe
> Assignee: Colin Patrick McCabe
> Attachments: HTRACE-164.002.patch
>
>
> htrace HRPC should use msgpack for serialization. Messages serialized using
> msgpack use less space on the wire and use less CPU time to encode. The CMP
> library allows us to include msgpack support easily in the htrace C client.
> There is also good Java and Golang support available.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
