The current auto instrumentation pipeline has the following problems:
- is not able to instrument top level apis of packages (iitm and ritm are able to do that)
- is platform dependent (tracing channels is a node api and requires polyfilling)
- requires annyoing workarounds because it is not powerful enough (cannot cleanly wrap functions and modify input output without ugly workarounds - a simple wrapping api would be better, instead of having to produce channels)
Proposed migration plan:
The current auto instrumentation pipeline has the following problems:
Proposed migration plan: