RE: Three GMPLS related IDs: draft-lin-ccamp-ipo-common-label-req uest-00.txt
Vishal Sharma <vishal@JasmineNetworks.com> Tue, 20 March 2001 11:00 UTC
Envelope-to: ccamp-data@psg.com
Delivery-date: Tue, 20 Mar 2001 03:03:05 -0800
Message-ID: <D72513E3A841D347A3545F0E2435E091E6E2DC@jasmine1.ad.jasminenetworks.com>
From: Vishal Sharma <vishal@JasmineNetworks.com>
To: "'mvissers@lucent.com'" <mvissers@lucent.com>
Cc: 'Zhi-Wei Lin' <zwlin@lucent.com>, "'xuyg@lucent.com'" <xuyg@lucent.com>, "'mpls@UU.NET'" <mpls@UU.NET>, "'ccamp@ops.ietf.org'" <ccamp@ops.ietf.org>, "'ip-optical@lists.bell-labs.com'" <ip-optical@lists.bell-labs.com>
Subject: RE: Three GMPLS related IDs: draft-lin-ccamp-ipo-common-label-req uest-00.txt
Date: Tue, 20 Mar 2001 03:00:39 -0800
MIME-Version: 1.0
Content-Type: text/plain; charset="iso-8859-1"
Folks, This is the last in the sequel. -Vishal Now on to the comments: i) First, a nit. The draft should be named "Common Label Request and Label Specification ...," since that is precisely the order in which it discusses things. ii) Also, (unless the plan was to make the wild goose chase of the implementer interesting!) at various points "Signaling Type" needs to be replaced with "Signal Type." The two don't mean the same thing. :-) iii) Section 2, Abstract. The statement describing the very first change needs to be expanded on somewhere in the rest of the text! It is not just a new _set_ of Signal Type structures, it is a completely new organization of what is there in GMPLS-SIG or in GMPLS-SIGEN, and that needs to be made very clear to the reader. The text should say that RGT, RNT, etc. have been all absorbed into the signal type. (Of course, it is an entirely different matter to say "why" that makes sense, but we'll come to that shortly.) iv) The final "signal type" discussed in Section 3.2.3 again breaks things into a signal type and a concatenation type. So it seems such a break-up is needed. What would be use then of eliminating RNC in GMPLS-SIG? If there is a valid technical reason for doing things this way, it should be clearly, succintly, and unambiguously articulated. I am afraid, in this trilogy of drafts that does not seem to have been done adequately. v) My major problem with this and its companion draft is that it proposes changes (some of which could even be good ones), BUT it does not provide a detailed enough justification for all of its changes. Nor does it provide the reader with any insight about why the encoding proposed in this draft is a good one, a better one, an easier one, or a sensible one. (Note: I am not implying that the remaining GMPLS drafts do a good job of this. I am here merely pointing to the _need_ to do so, in all drafts; in particular, in drafts that propose changing things (pl. see, for example, draft-mack-crane-gmpls-signaling-enhancements-00.txt and draft-bernstein-gmpls-optical-00.txt, which at least make an attempt to do so). This is especially useful to enable the community to appreciate the _technical merits_ of each proposal and is very^2 helpful to pick the right ones.) A lot of the community may not instantly know all of the technicalities of SONET/SDH or optical technologies, so it is all the more important to make a crystal clear _technical case_, that anyone can appreciate. vi) So why does it make sense to combine RGT and RNT and transparency, and signal type all into "signal type"? Why is that more helpful/better/cleaner than a form of encoding that treats these as separate entities? -- If we resort to one big coding space for signal types, it would be difficult to add more as more signal types come in, since all combinations of those signal types with RGT and RNC will have to be added. -- Further, what if an RNC or RGT option changes tomorrow? Now, we'd have to expand all existing signal types! If things were broken up, only the RGT or RNT codepoints would change. vii) The encoding of signal types in Section 3.2.4 allows for two ways to specify an STS-1, using value 5, X=0 or value 7, X=1. Was this intended? viii) Section 3.2.4, in the values for virtual concatenation (10-15), why does the draft stop at STS-3c-Xv? What about STS-12c-Xv, or higher? ix)Similarly, in the values for co-routed groups (23-27), why stop at STS-3c-X? What about STS-12c-X or higher? x) Page 8, para 3, the number of vendor specific codepoints is 12 it seems, not 13. xi) Section 4.2, the document spends many paragraphs trying to make a case apparently for why HOVC and LOVC labels are independent, and should have dedicated channel IDs. However, I cannot see how this follows from any of the preceding explanations. After a lot of words, the reader is suddenly confronted with the blanket statement " Therefore, HOVC.... labels. Each one will have a dedicated channel ID." xii) Again, in Section 4.2, the notation for the sub-STM-0/STM-1 signals is never introduced, even briefly. It is just thrown at the reader. Where does that come from? What is the background? The reader has no clue... xiii) In the same section, para 6, the last sentence says "At any moment, the complete aggregation bandwidth's multiplex structure must be defined to prevent alarms to be raised." Uh?! Say what? xiv) Section 4.2.3, please ... at least give an explanation of the notation used in the title, in plain English, somewhere. (Not everyone is an SDH/SONET expert, for crying out loud!) xv) Again, maybe I am missing something, but how does the current encoding allow to encode a signal like 3 STS-3? or 7 STS-3c-Xv? Or doesn't it? xvi) Also, it would be useful for the document to give references to all the G.707, G.708, etc. documents that it refers to throughout, together with pointers to how they may be obtained. (Let's make people's life a little bit easier, let's at least help them read, understand, and look at our drafts critically, instead of having them put the drafts aside in a huge pile, never to be read.)
- RE: Three GMPLS related IDs: draft-lin-ccamp-ipo-… Vishal Sharma