RE: [XCON] CPCP Requirement: Repeat times

[Michael Hammer <mhammer at cisco.com>]

Brian,

I may be a little confused by what is meant by policy and policy 
manipulation with respect to how that translates into resources being 
invoked and released.  My vision (delusion) as to how that might work might 
have to wait to see the actual protocol develop.  But, I suspect that your 
answers to my comments below will help clear this up for me.

Mike


At 01:33 PM 2/6/2004 -0500, Rosen, Brian wrote:
> > >End when last person leaves is the way many bridges work.
> >
> > Is that a technical justification?
> Actually, I think it means there is a requirement for that to be possible.
Agree.  Don't see that as not possible.


> >
> > >You might
> > >not like it, but that's the way they work.  Most bridges don't
> > >start if the convener is not there.  If he arrives, and then leaves,
> > >the conference is over.
> >
> > What is the meaning of conference policy then?  I thought
> > that it was an
> > authorization for something to take place.  If the conference
> > is authorized
> > to start and someone shows up, then it starts.  If it is
> > authorized to
> > stop, and the bridge resources are needed elsewhere, then it stops.
> CPCP has two elements, and they each have a role in policy.
> A particular implementation, or more properly, a particular administrative
> action, may restrict the range of possible policy requests the convener
> may ask for.   The bridge may not respond the way the convener would like
> it to.  It must be possible for the bridge to have local policy, and yet
> allow the convener to manipulate whatever freedoms the local policy
> permits.
I agree with this part.  The local "uber-policy" determines if the 
convener's policy request is accepted or rejected.


> You could have the bridge change the start/stop time itself
I don't follow this part.  It seems to confuse the role of the bridge in 
committing to the use of resources and the role of the convener in 
committing to pay for what he has asked for.  If it is the intent of the 
requester to terminate the conference upon his leaving, then that should 
be  an option of the request he makes to the bridge.

If he is requesting that the bridge terminate upon his departure, shouldn't 
that be an update to his policy?  Same for other options.  The bridge of 
course may have resource limitations, and thus there may be conditional 
authorizations to what is requested.  Where I think I have some confusion 
is that some of these options/authorizations are implicit not explicit.

> to implement
> its local policy, but that makes the function even more twisted.
> Better to have explicit mechanisms, and let the bridge implement, or permit
> a subset of them if it wishes.
>
> >
> > >It also doesn't work to manipulate stop time to end the
> > conference because
> > >its the bridge that wants to enforce its policy and not the convener.
> >
> > Who is the authority here?  Isnt' the bridge-providing
> > service supposed to
> > obey policy?
> As above, local policy must override.
I agree, in the sense that local policy drives convener policy that drives 
bridge resource provided/withheld.


> ....snip...
> > strategy, and an implementation detail.  It seems the
> > confusion here stems
> > from not separating:
> >
> > Is a conference authorized at the moment,
> > Is a bridge needed at the moment,
> > and should a bridge be reserved at the moment.
> I don't see it that way.  I see it as the bridge has local policy,
> the convener wants to exert whatever freedom the local policy permits,
> and the participants have to have some way to understand what is
> going to happen.
But, isn't the time at which the local policy interacts with the convener 
request independent of when a particular bridge resource gets assigned?  A 
convener request could be accepted so long as there are enough resources 
available/reserved in the pool, whether or not a resource is actually put 
to use.


> >
> > >We have to be explicit.
> >
> > Yes, particularly as to the semantics of policy.  What does
> > start time
> > mean?  What does stop time mean?  I keep seeing primary
> > meanings tangled up
> > with secondary and tertiary implementation implications.
> I've proposed that start time is the earliest time at which
> mixing starts.
I thought this was a policy?  This would mean that the start time would be 
the earliest time at which mixing is permitted to start.  Isn't this 
independent of whether a resource was invoked?  Note, it may be reserved 
un-occupied and billed for, but that may not be the most efficient use of 
resources.  (No statistical over-booking allowed.)


> I've proposed that stop time is interpreted by the stop time
> enumeration.  It could be a hard stop of mixing resources, or it
> could be only advisory.
Could this not be defined as the latest time at which mixing is permitted 
to be used, i.e. that the enumeration option is whether the stop time may 
get updated explicitly by the convener or another party or implicitly by a 
convener agent (could be the bridge)?


> >
> > >As to Cullen's objection, what you are worried about is start time,
> > >and not stop time.  By your definition, the conference never started.
> > >If it did start, then it ends when whoever showed up leaves
> > (if that's
> > >what the conference policy was set to), or when the convener leaves,
> > >which is surely the right thing.  I don't know how to fix
> > your problem
> > >actually.   Maybe some kind of action that resets start.
> >
> > The bridge may need to be released and re-captured, but that
> > does not mean
> > the conference is re-started.
> Possibly a minor semantic argument.  I think we both agree that
> the scheduled conference did not take place.
Are we talking when a convener was authorized to invoke a conference 
resource or accounting for when resources were actually used?
I believe that some conference services still charge as the resources where 
reserved and represent an opportunity cost.

> What is the
> difference between "release and recapture" and "reset start"?
One appears to be a new conference instance, while the other appears to be 
a modification of an existing conference.


> >
> > >   The problem is,
> > >that would have to be a convener action; you wouldn't want a
> > participant
> > >to invoke it.
> > >
> > >Also, Petri, you can't implement a "Meet Me" with
> > start/stop.  MeetMes don't
> > >have start and stop times.
> >
> > You are asserting what the group is attempting to define.
> Well, we can twist start and stop, but to any person, a meet me, by
> definition, does not have start and stop times.  Could we implement it
> by twisting start/stop?  I guess, but I don't think its a good idea.
>
> >
> > >   They start whenever the convener arrives,
> >
> > What if the convener never arrives, but the rest of the group
> > does?  To say
> > it never starts seems rather limiting in an unnecessary way.
> That's the way Meet Me (notice the word "Me") services are defined.
I should note that the term "meet me" has been used in a more broad way to 
mean any pre-scheduled conference for which a number (address) has be setup 
and to which users are asked to call into at some future time.  You seem to 
have something more specific in mind.  I have no problem with a specific, 
maybe trademarked, service to be more narrowly defined, but see no reason 
to limit other possibilities.

> I happen to think its an excellent service.  I would not want anyone
> to be able to use my Meet Me conference if I was not there.
> I am charged for it.  If 4 people know my MeetMe URI, they can decide
> to have a conference (on my nickle) any time they want if there is
> not a restriction that convener starts.  That is precisely why MeetMe
> works the way it does.  Its not a current technology limitation.
I have not problem with this being possible, just don't think CPCP should 
be restricted for being used in other ways.  Protocol should be generally 
usable for many types of services.

> I have no problem at all in allowing CPCP to facilitate an ad-hoc
> conference, which is what I think you are looking for.  Ad-hoc is
> a conference starts either by promoting a one on one to a three way,
> or by the first person connecting to the conference, and it ends when
> the last person leaves (or for some, when the number of participants
> drops below 3).  Ad-hoc is also a useful service, but usually we use
> ad-hoc internally where charging issues are simpler.
> >
> > >end
> > >when he leaves (usually),
> >
> > This again is not necessary.  A convener (payer) may want to
> > start the
> > conference, but not stay for the full discussion.  No reason
> > the conference
> > can not continue without that person, if the policy says it can.
> Agree, but I want you to explicitly say that, and I want local policy
> to be able to permit you to say that, or deny you the ability to say that.
OK


> >
> > >but restart when he arrives again with out regard
> > >to start and stop.
> >
> > A new start and stop time is a new conference, perhaps an ad hoc
> > one.  There is no reason the conference policy would restrict
> > the convener
> > from scheduling multiple conferences, unless there is other
> > local policy
> > that over-rides the ability to do so.  But, that again is
> > implementation.
> Well, a new start/stop MAY be a new conference, or it could be a reschedule
> of an existing one.  Take a scenario where a couple of participants show,
> the convener doesn't, and the participants leave.
Hopefully, the convener could show up late and the conference is still 
active and resources get invoked for him and whoever else shows up late.

> The convener pushes
> start and stop an hour.  If its a new conference, there is a new conference
> URI.
> If its a reschedule, then its the same conference, and the same URI.  I
> think
> that is preferable.
So long as the convener shows up within the hour it was originally 
scheduled, could it be neither a new conference nor a reschedule?  Suppose 
he shows 20 minutes late, but only needs 40 minutes to get the work done, 
why should any policy change take place?


> Local policy, as I advocate, is indeed a controlling factor.  However, to be
> successful, the convener and the participants have to know what the local
> policy is, so that their expectations are met.  That means CPCP has to have
> ways of the participants including the convener to learning the local
> policy,
> which means the language has to have the ability to express it.  It's not
> just implementation.
>
> >
> > I hope the tone of the above does not sound harsh.  I just
> > think it would
> > be useful to separate the definition of "what" the conference
> > is from the
> > definition of "how" it is to be supported.
> Confused about "what a conference is".  I think we are all working on the
> "how it is to be supported".  I don't even want to think about a definition
> of what it is.  For example, if I play music on hold to you while you are
> waiting for a conference to start, is that part of the conference?
That sounds like a service description which may be implementation dependent.

> You got
> there with the conference URI, you can probably manipulate some resources

_______________________________________________
XCON mailing list
XCON@ietf.org
https://www1.ietf.org/mailman/listinfo/xcon