* Tim van Dijen <
tvd...@gmail.com> [2021-09-06 14:35]:
> The only question I have is; is my assumption correct that this
> happens in an IDP-first flow? Because I would have expected the IDP
> to have refused an AuthnRequest addressed to another entityID..
That's a bit of a red herring, I think:
It's SP initiated, but with unusual deployment specifics at play:
The IDP is a (supposedly mostly identical) *test* instance of the real
IDP and the IDP operator has used his local workstation's hosts file
to point his web browser to the test IDP instance for the prod IDP's
host name (instead of consulting the DNS which points to the prod
IDP's IP address).
As long as your test IDP has identical entityID and key material as
the prod IDP that trick allows you to do end-to-end tests with
arbitrary SPs. The mistake here was to give the test IDP its own
entityID, differing from the prod IDP the SP had metadata about.
So I guess for the sake of the argument it could be said to be IDP-initiated.
Best,
-peter