[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Freeform labels




On May 11, 2006, at 10:29 AM, jcucchiara@mindspring.com wrote:



-----Original Message-----
From: Tom Petch <nwnetworks@dial.pipex.com>
Sent: May 9, 2006 3:43 AM
To: Adrian Farrel <adrian@olddog.co.uk>, ccamp@ops.ietf.org, tnadeau@cisco.com
Cc: joan cucchiara <jcucchiara@mindspring.com>
Subject: Freeform labels

Hi Tom,


I am confused about freeform labels. I am not familiar with them from other
contexts and so am not clear how significant they are but ...

(draft-ietf-ccamp-gmpls-)tc-mib under GmplsFreeformLabelTC references RFC3471 which would be a good place to read about them except I can see no mention of them there. They appear in tc-mib, lsr-mib and te-mib - they are used as worked examples which is what attracted my attention - but I cannot find them anywhere else in the (G)MPLS library. I think that more description is needed somewhere.

The GmplsFreeformLabelTC is also used in the GMPLS-LABEL-STD-MIB (gmplsLabelFreeform in the
gmplsLabelTable).  I believe given the purpose of this particular
MIB table and also within the context of the other objects in this
table, that the intent of GmplsFreeformLabelTC is fairly well described.


Why does GmplsFreeformLabelTC give size limits of 0 and 64? Can they really be 0-length? I have no problem with an SMI object having zero length as long as the entity being modeled by it can have zero length but is that meaningful here? And why 64? ok, a lot more than 32-bit but 64 sounds arbitrary (in which case, I think that worth saying); or does 64 octet relate to some other well known
address formats (IPv6 anyone?).


The length of 64 was thought by the authors to be more than enough for the purpose of this label. I also questioned this during the MIB Dr. review, but currently label
lengths in the MPLS/GMPLS wgs are 4 octets, so 64 seems reasonable.

0 length is used if this object does not contain a label, so it is not the actual label
that is 0 length, but when the object has not label.


When GmplsFreeformLabelTC says 'not defined in any RFC', I think it sums it up
too well:-(


Yes, this is a MIB convention only.

So the use of gmplsFreeformLabel(3) everywhere except in the formal definition where it appears (I assume) as gmplsFreeformGeneralizedLabel(3) seems about
right as does the descripttion


Good catch. I probably should have caught this during the MIB Dr. reviews.
Tom N. could you change this during the IESG/RFC reviews.

	Will do.



-Joan


"The label is any form of label
encoded as an OCTET STRING using
the Textual Convention GmplsFreeformLabel"

Well no, the Textual Convention is GmplsFreeformLabelTC and I would rather say that the label is encoded using the Textual Convention GmplsFreeformLabelTC
(even though that is a redefinition of an OCTET STRING).

I think more should be said somewhere, ideally in RFC3471, but, more
realistically, in tc-mib.

Tom Petch