Open Stack

A few days ago, I was talking to Rossella, about the need of a RFE for 
this [1],
in this case, it's an internal library/nicety for neutron development. 
It's not a feature
from the user point of view. And it's not a very big change.

So, we thought that starting by a simple devref to agree on, and 
continuing by adding
to the same patch the code that goes with the devref may be enough. Kyle 
agreed with
the idea, but we wanted to share it on list.

The advantages of this approach are interesting:

1) We end up with a devref we can keep on tree directly and that servers 
as developer
      documentation.

2) We save time and simplify the process by doing it all in the same place.

I agree that RFEs/specs are valuable in the context of the wider 
openstack project, but
are better for user facing functionality or big design changes.

TL;DR

About [1], it's something to be consumed by the QoS design, but made with
reusability in mind.

We expect to reuse it for security groups in the future, improving the 
design and
optimizing the way agents consume messages in the end thanks to
object-type + id  directed  fan-outs.
(may be such second step would deserve a RFE + little spec to be evaluated).

[1] https://review.openstack.org/#/c/190635/ (Generic RPC mechanism 
which could be reused)

Best,
Miguel Ángel Ajo