> -----Original Message-----
> From: [email protected] [mailto:[email protected]] On 

> My understanding is that it labels a flow. If the right way 
> to carry a flow is distributed across two network paths, so 
> be it. BTW, a "flow" has many definitions; the type of flow 
> that Tony was mentioning this morning is a single application 
> session; to an ISP, a flow is a stream of traffic from an 
> ingress to an egress, or perhaps the stream of traffic from 
> an ingress to an egress via a path. This means that the IPv6 
> Working Group wants the flow label to be a hash that helps 
> the network find status for a specific session's or set of 
> sessions' shared' state, while to an ISP it wants to be 
> independent of the source address - associated with the 
> ingress router.
> 
> 
> RFC 1710, the original specification for what later because 
> IPv6, said of the flow label:
> 
>    The Flow Label field in the SIPP header may be used by a host to
>    label those packets for which it requests special handling by SIPP
>    routers, such as non-default quality of service or "real-time"
>    service.  This labeling is important in order to support 
> applications
>    which require some degree of consistent throughput, delay, and/or
>    jitter.  The Flow Label is a 28-bit field, internally 
> structured into
>    three subfields as follows:
> 
> RFC 1752 agreed:
> 
>    * Flow Label - This field may be used by a host to label those
>      packets for which it is requesting special handling by routers
>      within a network, such as non-default quality of service 
> or "real-
>      time" service. (28-bit field)
> 
> RFC 2460 reads:
> 
>       o  Flow Labeling Capability
> 
>          A new capability is added to enable the labeling of packets
>          belonging to particular traffic "flows" for which the sender
>          requests special handling, such as non-default quality of
>          service or "real-time" service.

But didn't we determine in a long thread that such uses by routers were not 
very robust, because there is no header checksum to protect that flow label?

If I want to use flow label end-to-end, any rule that says that flow label is 
immutable unless set to zero becomes ambiguous. At the destination host, I 
would have no idea whether the source host set the label, or whether a router 
along the path set the label.

But I understand that is how they ARE being used now, so I guess the point 
becomes moot.

Bert
--------------------------------------------------------------------
IETF IPv6 working group mailing list
[email protected]
Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------

Reply via email to