Re: Fw: more changes to the base draft
Avri Doria <[email protected]> Sat, 6 Dec 2003 10:50:14 +0900
| Newsgroups | gmane.ietf.gsmp |
|---|---|
| Message-ID | <[email protected]> |
On l=F6rdag, dec 6, 2003, at 10:03 Asia/Seoul, Khosravi, Hormuzd M = wrote: > Hi Avri > > I had one question on this.. > Technically speaking GRMP seems quite different from GSMP v3 > to me. How could you possibly merge these protocols ? > I must be missing something here. > > > Thanks > Hormuzd > > The structure of the two protocols is very similar in that GRMP uses=20 similar protocol protocol mechanisms, in fact GRMP imports much of its=20= protocol structure from GSMP almost completely. What is different,=20 primary, is the way the information is expressed. So yes, it is true=20 that the messages for setting state for GMPLS objects is different then=20= the way GRMP sets state for the LFBs, but this can be accommodated by=20 introducing a new message type, or two, for LFBs. I am also assuming,=20= though I have not worked it out yet, that much of the technology=20 specific changes that have been made to support various optical=20 technologies could also be used. As I indicated in my first message to Weiming, we would need to do some=20= work to see if the merge is really feasible, but on the basis of my=20 initial review, I don't think it would be impossible. My initial=20 reasons when the ForCES group first started out for not really being=20 interested in having GSMP adapted to support ForCES was that it could=20 complicate the simple structure of GSMP. Now that we are almost=20 finished with the rework of GSMP, it would be V3.1, for GMPLS, the=20 structure has gotten at once more flexible and more complex and I think=20= the merge could be done with a minimum of pain. The advantage I see in this is that a controller that could use GSMP=20 for controlling GMPLS switches might, with the addition of some=20 specific GRMP-like modules could also support IP forwarders. This=20 could be a very strong argument for the folks in Asia, and perhaps=20 elsewhere, who are already developing and using GSMP for switch=20 control. I guess the point is that now that the GSMP base has been=20 generalized to support any of the varieties of GMPLS switch, it is more=20= like other general carrier protocols (except that it is a vertical=20 protocol) - mostly what you have to do is define a payload format for=20 the content you wish to carry. But, this is a very high level analysis, and if we were to do this, and=20= that would involve ForCES and AD consultation and charter update etc,=20 we would need to put together a design team to work out the details for=20= real. a.