[Kea-users] Client classification

[Sutherland, Rob]

Here's a real example of a test (currently in production with the "old" DHCP server):

  {
      "name": "phones",
      "test": "(substring(pkt4.mac,0,3) == 0x08000f) or (substring(pkt4.mac,0,3) == 0x000d65) or (substring(pkt4.mac,0,3) == 0x000413) or (substring(pkt4.mac,0,3) == 0x0050c2) or (substring(pkt4.mac,0,3) == 0x0004f2) or (substring(pkt4.mac,0,3) == 0x64167f)"
  }

It's a silly but effective test to determine which devices are to be considered as phones. In Kea 1.3.0, I need to use this as the "not":

  {
      "name": "not_phones",
      "test": "not (substring(pkt4.mac,0,3) == 0x08000f) and not (substring(pkt4.mac,0,3) == 0x000d65) and not (substring(pkt4.mac,0,3) == 0x000413) and not (substring(pkt4.mac,0,3) == 0x0050c2) and not (substring(pkt4.mac,0,3) == 0x0004f2) and not (substring(pkt4.mac,0,3) == 0x64167f)"
  }

Rob

-----Original Message-----
From: Francis Dupont [mailto:fdupont at isc.org] 
Sent: Thursday, November 02, 2017 6:21 PM
To: Sutherland, Rob
Cc: kea-users at lists.isc.org
Subject: Re: [Kea-users] Client classification

"Sutherland, Rob" writes:
> I have a quick question: if I have a client class named, say, "foo", 
> how do I define a class for everything NOT in "foo" (without negating 
> the entire test defined in "foo")?

=> not yet but it is in my TODO list. BTW if you have a concrete example it will help for the documentation (real world examples are always better) and to convince my colleagues there is an important item which should be do in Kea 1.4 with a good priority. Note there should be a design phase as it is not the only thing which can (should!) be improved in client classification and it is not trivial too (dependency loops must be detected).

Thanks

Francis Dupont <fdupont at isc.org>