What should happen if sampling decision or parentId is an invalid value?

[srjames90]

If debug flag is invalid, the spec says just ignore it and move on, but the spec doesn't say what to do if it parentId or sampling decisions are invalid. Do we do the same thing, or just return the context? Zipkin-go and Zipkin-php seem to just return the context.

[codefromthecrypt]

Sorry you are right we do not define in this spec malformed state handling. Things are mature enough that we can, though I wouldn't whip libraries in reverse too much as this sort of thing isn't always specified anyway. MAY not MUST in other words. For example, brave is pretty strict and all things except some known old client misbehavior we "restart" aka start a new trace https://github.com/openzipkin/brave/blob/master/brave/src/main/java/brave/propagation/B3Propagation.java There are many libraries implementing B3 so we'll want a ping list if you feel up to making a chart. Meanwhile we can say something like.. implementations are not required to handle malformed trace contexts, but doing so not necessarily a bug provided X (where X is complex because accepting w/o span ID definitely would not be encouraged). The problem is getting into the nuance will tempt people to handle malformed contexts which are usually programming bugs better fixed upstream these days.

…

On Fri, Jul 31, 2020 at 7:54 AM srjames90 ***@***.***> wrote: If debug flag is invalid, the spec says just ignore it and move on, but the spec doesn't say what to do if it parentId or sampling decisions are invalid. Do we do the same thing, or just return the context? Zipkin-go and Zipkin-php seem to just return the context. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#41>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAAPVV7QQCN56ETZBFLYT3LR6IB45ANCNFSM4PO55NDQ> .