El 25/05/2011 4:22, Fatai Zhang escribió:
Hi Ramon,
If I understand you correctly, do you meant that you prefer "only
the VENDOR-CONSTRAINT object" to "one extra TLV"?
Hello Fatai, all,
I'm afraid not (sorry for not having expressed myself clearly). In the
case of vendor constraints, I think, without a strong opinion, that I
would rather have vendor-specific TLVs that apply to a given object,
rather than a top-level object. The reasons for this are:
* It seems more granular and clearly identifies which object is
constrained. Seems less opaque :)
* The default behavior as current RFC is to ignore the TLV if not
known/supported. This may require a flags field within the TLV to
specify that the constrain was indeed processed.
A good question is whether a single TLV "vendor_tlv" would do, or there
is a reason to have more grained tlvs
There are some drawbacks of course:
* There are objects that were defined without TLVs. With a strict
interpretation of rfc 5440, it would not be possible to constrain such
objects.
* The constraint (semantics) involved in the TLV has to somehow relate
to the object it is attached to. If there is no such object, the use of
TLVs seems less flexible and an object would be needed.
In short, there seem to be use cases for both. Preferably, use vendor tlvs.
Thank you and best regards
R.
_______________________________________________
Pce mailing list
Pce@ietf.org
https://www.ietf.org/mailman/listinfo/pce