The MPLS WG Archive

[Erblichs <erblichs@earthlink.net>]

Here a re a few quick comments:

	1) 
	Abstract:
	"10s of milliseconds"

	My assumption is that you are referring to
	a SONETs capability of detecting a Link status
	change in approx 10ms.. Additional ms delays
	are expected.

	When OSPFs Router-dead-Interval's of 10s of
	seconds (ethernet) probably precludes the 
	use of this extension,

	Thus I would suggest a re-wording of
	"These mechanisms enable"

	enables approximately a 10ms SONET's detection
	of a Link down status to trigger subsequent
	redirection of traffic onto temporary alternative
	LSPs known as backup LSPs.

	Additionally...

	All the methods specified within assume that
	backup LSPs can be formed for redirection of
	traffic due to a link-failure using alternate
	paths.

	Questions:

		Should backup LSPs accept lower QOS
		capabilities when equal QOS capabilities
		do not exist?

		What is the order of setting up backup and
		non-backup LSP tunnels?

		If we setup backups first. do we route that
		way until the non-backups are formed? Would
		we assume that it would take less time to
		form the backup LSPs?

		What is the lifetime of a backup LSP?

		How do we withdraw a backup LSP?

	2) One-to-one backup

	In general RSVP has implied scalability concerns
	vs diff-services. With the additional load of
	"One-to-one backup", are those concerns increased?

	I did not see any mention of scalability comparisons
	with this and "Facility backup"? Would the nomenclature
	of "1-2-1" be correctly understood? "1-to-1"?

	3) NITs..

	Should the figures in 5.1 have the ONLY value
	   specified in the field?

	Ex: Class-Num   -> Class-Num = 205

	4) 5.1
	
	Their are two possibly conflicting statements
	dealing with "Class-Num" and "Unknown C-types"
	on how the object should be treated..

	What do we do if Class-Num specifies that a
	"should ignore it and pass if forword unchanged"
	and a C-types says soemthing else? Which has
	precedence?

	NIT, I assume that "pass if forward unchanged"
	should be "pass IT forward unchanged".. If I am
	wrong, you want a node that doesn't understand
	a object to possibly change it, then maybe forward
	it........

	Mitchell Erblich

_______________________________________________
mpls mailing list
mpls@lists.ietf.org
https://www1.ietf.org/mailman/listinfo/mpls