Closed pavolloffay closed 5 years ago
Given that we already have the method in http.method
, do we really need this? Seems like it just makes it more difficult to get all traces for a given endpoint, regardless of the method. If the method is important, just add the tag to the search?
This has been proposed in MP-OT spec. Unfortunately, we have to align with the spec and TCK. For example post operation might have a different path and performance characteristics, the OT spec advises to use different operation names for different operations.
The workaround is to use custom operation name provider or override the operation name with @Traced
annotation for given endpoints.
Sorry for commenting on a PR that was merge a long time ago. I'm curious to know more about the proposal in the MT-OT spec. As @wjoel mentions, having http.method
as a tag may achieve the same goal, but I may be missing some context on the reasons why this was proposed.
I can see the MT-OT spec on the creation of spans for incoming HTTP requests, but I don't think OpenTracing specifies how HTTP server spans naming, and this seems to clash with the OpenTelemetry spec on HTTP server semantics.
Do you see this naming convention as something being more standardised in OpenTelemetry for any HTTP server implementation?
Signed-off-by: Pavol Loffay ploffay@redhat.com
The build will probably fail MP 1.2.1 was just published. I will restart CI and merge later.