Comments inline.
Michael
"Rosen, Brian" wrote:
I think you missed one point: "How does a NON-redundant MG upgrade to a new version of the protocol?"
It comes down, reboots, and comes up again - it's a service affecting change. AFAIK, that's the only way today.
<MB> Yes, if the MG is really Non-redundant there doesn't seem to be an option.
Can the MG Failover to itself? Will this cause any problems in the MGC?
Since that's not a defined option, some MGCs might have a problem.
<MB> While there is no existing text specifically saying that this is allowed., there also is none that disallows this behavior. It seems to me that the MGC should just accept the "new" address for the "secondary" MG and go on. This would be used when an MG and its mate hide behind a single address for the MGC <> MG control association. I would think that this could easily be added to the implementor's guide and I don't think that it should cause problems.
To allow this requires only a few changes to the text in RFC3015 as follows: Section 7.2.8 ServiceChange Modify text: "6) Failover - sent from MG to MGC to indicate the primary MG is out of service and a secondary is taking over, or to indicate a version negotiation from the MG." Add a Reason: "Version Negotiation" or a more generic "MG Directed Change". Add text: "The MG may initiate a version negotiation with a ServiceChange Command specifying the "Root" for the TerminationID, ServiceChangeMethod equal to Failover and ServiceChangeReason equal to "Version Negotiation"."
I think that's a scope add, which I for one am reluctant to do. If there was a real problem with something we already had, that's one thing, but a real feature add, I'm less thrilled about. I'll admit some of the things we have talked about stray at least close, if not over such a line. Here, I think an MG that can do a non-service affecting upgrade which is not redundant is pretty unusual these days - I don't mean to discourage folks from doing that, but I also don't feel we need changes in semantics (and some additional syntax, albeit some IANA registrations) to make it happen, it can wait for V2.
<MB> I'm suppose I'd agree that this propoal is adding a new feature as it would require a new, IANA registered reason and the associated behavior. I'd say it's somewhat unfortunate because basically, I don't see problems with the proposed approach. There is one thing about the approach that bothers me which we might consider for V2. That is, the use of Failover in this instance seems inappropriate. The name just seems to imply that a software upgrade is a failure scenario. I would suggest that it would be nicer to enhance the definition of Handoff rather than Failover since handoff implies a more graceful transition. Given the current definitions of Handoff and Failover in version 1, we do have to live with Failover for now.
Cheers, Paul
"Rosen, Brian" wrote:
It's unlikely that we will do anything that adds new features by edits to the implementor's guide. It's reasonable to have explanatory text in such a document that is not normative.
I think you failover, upgrade, and failover back with this version of the protocol.
Brian
-----Original Message----- From: Paul Rheaume [mailto:paul.rheaume@alcatel.com] Sent: Thursday, February 22, 2001 1:00 PM To: ITU-SG16 Mail List Cc: megaco@fore.com Subject: Re: Protocol version re-negotiation
Folks,
Where is the appropriate list to discuss this issue?
We feel that version negotiation and the version upgrade process is not well specified and needs to be discussed and properly
defined at the
latest in version 2 of the specification. That means
looking at this
very soon.
I think this issue belongs to the next version of the
Implementor's
Guide.
Regards, Paul
Nancy Devin wrote:
How does a non-redundant MG upgrade to a new Megaco version without affecting service? How does a redundant MG upgrade to a new Megaco version without affecting service?
It is mentionned in RFC3015 that Protocol version negotiation can be done at Restart, Failover, and Handoff serviceChanges. It is not mentionned explicitly how version protocol re-negotiation is done after the MG has registered with the MGC. in both situation
This could be done in at least two different ways as listed bellow.
1- Protocol version negotiation could be initiated using the Failover serviceChange. A Failover serviceChange with a protocol version would indicate that the MG has been configured with a new H.248 protocol version. Sent from the MG to the MGC, it indicates that the MG has been configured with a new H.248 protocol version. This serviceChange shall not indicate Failure of the working MG. This would require that for non-redundant MGs, both the primary IP address and the secondary IP address be the same.
2- Protocol version negotiation could be intiated using a new serviceChange method: VersionNegotiation (Method 7 of the serviceChange command as defined in RFC3015) This serviceChange would be sent from the MG to the MGC to initiate protocol version initiation. It indicates that the MG has been configured with a new H.248 protocol version. Sent from the MGC to the MG, it indicates that the MGC has been configured with a new H.248 protocol version.
Comments please. Nancy