Closed codefromthecrypt closed 8 years ago
I think it might be valuable to be able to trace every call individually but also be able to from an application trigger some other logic. Consider the case of a page being open for hours with lots of short api calls creating a huge unwieldy trace view
Try looking at http://m.finn.no/ on the network pane, leave it open for a minute, watch things going on... it's quite a circus! I don't really think it would make sense to collapse all these requests together into the same trace. In the extreme case, you would have a single-page webapp without any site reloads at all. Then, 60 minutes of user activity would be collapsed into 1 single Zipkin trace. In my head, that doesn't sound quite right...
I think it is probably best for the app to trace itself in order to collapse things like this. Once we have a javascript tracer, we could do this in zipkin-ui itself, for example.
when tracing zipkin, I noticed that each request for the search page is in a different trace id. While this adds value (ex predictable trace ids), I suspect we'd get more if they were in the same trace. Is this feasible?