Re: [Dime] [dime] #45: Why is a validity duration of 0 disallowed?

My understanding was that e.g. if the reporting node has requested 30% reduction but now detects that it needs a 50% reduction, it just sends a new OLR with 50% and that will implicitly indicate that the previous OLR with 30% is no longer valid. This is a valid way to make the 30% OLR invalid; no need to first send an zero validity time OLR to make the 30% OLR invalid. Any (correctly coded) fresh OLR makes a previous OLR invalid.

The second sentence (at least in my understanding) is ambiguous: is it the OLR that contains a zero percentage that should not be regarded invalid or is it the previously received OLR?

I have no strong view whether or not to allow zero validity and/or zero percentage, but the proposed text is not understandable/acceptable.

Ulrich

From: DiME [mailto:dime-bounces@ietf.org] On Behalf Of ext Steve Donovan
Sent: Monday, February 24, 2014 10:33 PM
To: dime@ietf.org
Subject: Re: [Dime] [dime] #45: Why is a validity duration of 0 disallowed?

I think the proposal here is the following:

A reporting node communicates that an overload report is no longer valid by sending an OLR with a Validity-Period AVP with a value of zero.  This is the only way for a reporting node to indicate that an overload report is no longer valid.  For instance, setting the reduction-percentage to zero does not make the overload report invalid.

Do we have agreement on this?

Steve
On 2/17/14 12:44 PM, Ben Campbell wrote:

On Feb 17, 2014, at 12:08 PM, lionel.morand@orange.com<mailto:lionel.morand@orange.com> wrote:

I would like to highlight that my proposal is not related to the default Reduction-percentage algo.

Whatever the ago used, I think that an OLR can only be sent for two purposes:

- Send an indication of overload situation

- Send an indication of end of the overload situation

For me, providing both indications in the same message would be a non-sense. And I don't understand why the validity period would prevail in that case.

So here's the question: Let's assume a reporting node sends you an OLR using the "loss" algorithm, a reduction-percentage of 50%, and a Validity-Duration of 10 minutes. 5 minutes later, it sends you a new OLR with an invalid reduction percentage and a Validity-Duration of 0.

What's the current overload state?

What if the second OLR used the "rate" algorithm, with a valid value, but still had a validity duration of 0. Does that change anything?

I am arguing that, in both cases, the second OLR ends the overload period.

I'm actually okay if we have a normative statement to the effect that the reporting node MUST NOT send an invalid reduction percentage, and that a reacting node MUST ignore an OLR with an invalid value _except_ for the case of a validity-duration set to zero.

This all comes down to how much guessing we are willing to do about the intent of the sender when one receives an invalid OLR. I think it's reasonable to guess that a zero validity duration means to end an overload condition, without regard to the rest of the message. I _don't_ think it's reasonable to try to guess what the sender meant with an invalid reduction percentage, when the validity period is non-zero.

Another, perhaps simpler, approach would be to assert that the reception of _any_ invalid OLR automatically ends any existing overload condition. (I'm not necessarily arguing for that--it's just something to think about.)

Now, if what is argued by Ben and supported by JJ and Steve is clear for the rest of the group, I will not fight.

... but the fighting is the fun part :-)

_______________________________________________

DiME mailing list

DiME@ietf.org<mailto:DiME@ietf.org>

https://www.ietf.org/mailman/listinfo/dime