Closed dherman closed 11 years ago
The two different url
fields don't meaningfully distinguish the two requests. In the same way that two syntactic import
declarations from two different modules named "x1"
and "x2"
will get turned into one request, these two will become one request, even thought the referring url is different, and thus either both will succeed or both fail.
There is one way that these could be different, which is if the normalize hook for System
produces different module names for these two, based on the url (or based on Math.random()
).
@dherman points out the the rest of the pipeline (after normalize
) will see the referrer information from the first request (again, assuming that normalize
produces the same thing for both calls).
Right, this makes perfect sense to me now. I wrote that comment before I understood what options.url was actually for. Bogus issue.
What should happen when:
(Note the different urls.)
If we see the registry purely as a cache, we should really get two competing loads here, and whichever one finishes later gets an error... hmm. Perhaps a synchronous error on the second
System.import()
call would be better.