Closed boucadair closed 11 months ago
Joe: "I would emphasize this in the document to be clear what hardware ramifications might exist and what operational tradeoffs one would consider."
OK. Reordered the options and added a pointer to the new implem section.
From Joe:
"There was a question asked about needing deep packet inspection to effectively implement this. Quifang said yes, but I don't think it would be necessary. If the controller maintained the state and knew who the user was through other means (e.g., AAA/dot1x), it could program the network elements with standard ACL tuple data (i.e., traditional ACLs) dynamically, thus not putting more logic onto the devices or into the hardware. This was similar to a past comment of mine, and I think the document text addresses this.
It's not to say an implementor couldn't do something fancier within the network, but I don't think additional capabilities are required to make this work. "