random spitballing

[Raoul Duke]

per request,

if every step of the distributed system annotated each processing step then there would be a journal of how far along in the process something advanced before erroring out, or completing. it could also log all the times from individual local processing. 

not a log in another db but data appended to the request as it flows through local processes. maybe with a slightly lazy backing store of that for forensics when a process outright dies dead. 

if the system knows through introspection what the expected tree of possible/defined/allowed data/control flows is for a given request then the request journal could be compared against it. 

helpful in debugging, maintenance, sanity checking, performance analysis, security checking, etc. 

and we should certainly not be allowed any syntax that allows for any kind of non-asynchronous assuming.