[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [MEXT] [WGLC] draft-ietf-mext-flow-binding-03
Folks,
Please find below my _individual_ WGLC comments:
As we've discussed in the past I think the terminology is misleading and we had agreed to fix it. The draft hasn't been updated since then, however, so I'm repeating my comments in the WGLC.
The current flow description can catches traffic to many destination, e.g., HTTP traffic to any destination, yet the current definition of flow talks about packet exchanged between two nodes:
Flow: A flow is identified as a set of data packets that are
exchanged between two nodes and match a given flow description
I think it is fair to say that the basic object that this specification manipulates is a "bunch of flows" rather than a flow. So I think we should make the following easy change and renaming in our definitions and terminology, it will make the specification more consistent and easier to read without entailing changes to the protocol mechanism:
Flow:
(something ala RFC 2460, e.g.) a sequence of packets
sent from a particular source to a particular (unicast
or multicast) destination for which a MN desires
special handling by intervening MIPv6 peers, being it
the HA or the CN.
Flow Description -> Traffic Description/Selector (I don't care which one we pick, traffic selector is used by IPsec already with exactly the same meaning that here):
a piece of information that identifies one or more flows.
The identification is based on various fields of the
network and transport layer headers, such as the source
address, the destination port, the protocol number, the
DSCP value, etc.
Flow Identification -> Traffic Identification
Flow Identifier -> Traffic Identifier:
an unsigned integer used as a reference to a particular
Traffic Description/Selector.
Flow Binding -> Traffic Binding:
an association between a traffic description/selector and
an action (Forward or Discard). An IP packet that matches
the traffic description/selector will be processed according
to the action.
Note above, the Flow identifier doesn't need to be included in the definition of a binding. Also I prefer to use "Forward" rather than "n-cast", since Forward is clear and does not necessarily imply that the packet is only forwarded in one direction.
--julien