[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [ippm] draft-ietf-ippm-duplicate
At 04:17 AM 4/18/2007, Henk Uijterwaal wrote:
This looks fine too, but as far as I can see, it is exactly the same
as for a duplicate IP packet outcome.
Yes, the "set-up" conditions are the same for both. In fact,
you can't have a replicated packet without subsequently observing
a duplicate.
And then, for DIPPO:
> The egress reference event at MP_i for a duplicate packet
occurs subsequently to at least one other corresponding egress reference
event for the original packet (usually also at MP_i ).
I can understand this too, it seems to say that the DIPPO occurs after
the first copy arrived. That makes sense. But then for RIPPO:
The egress reference event at MP_i for a replicated packet is the first for
the original packet and occurs prior to at least one other egress reference
event for a duplicate packet (usually also at MP_i ).
This seems to say that the RIPPO occurs before the reference event.
It occurs before a DIPPO on egress, and can only be determined after a
DIPPO has occurred. Note that "least one other egress reference event for
a duplicate packet" or DIPPO is a necessary condition in the definition.
But how can one tell? A packet is sent and received, how can the receiver
tell that a second copy will arrive until that copy actually arrives?
The RIPPO can only be determined after a DIPPO. There has to be some
memory in the receiver to determine a DIPPO, and the packet that the
DIPPO corresponds to would be designated RIPPO.
But then what is the difference w.r.t. DIPPO: in both cases 1 packet
is sent and 2 copies arrive. (And the first copy is presumably used
by the recipient, anything else doesn't make sense).
The first of the copies is designated as RIPPO.
Maybe it would help to define a Non-Replicated IP Packet Outcome,
but I thought that wasn't necessary. All successful egress packet
outcomes are non-RIPPO initially. If there is a DIPPO, then there must
be a corresponding RIPPO to designate.
Multiple DIPPOs could come from one RIPPO,
or the same number of RIPPOs, or some other combination.
So distinguishing and counting the RIPPOs helps to determine
whether 50 duplicate packets came from one packet or
from 50 RIPPOs.
Another way to distinguish these cases would be to count the
first DIPPO, e.g., DIPPO_1, separate from subsequent DIPPOs that
correspond to the same packet, DIPPO_2, DIPPO_3, ... DIPPO_n
Or is there another subtle difference between DIPPO and RIPPO that I've
missed?
Now that I've explained it further, maybe you can suggest a way
to modify the definition of RIPPO so that it is clear?
I'm open to suggestions.
Al
_______________________________________________
ippm mailing list
ippm at ietf.org
https://www1.ietf.org/mailman/listinfo/ippm