[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Feed-back on <draft-ash-mpls-diffserv-te-class-types-00.txt>
Jerry, Waisum,
Let me separate two aspects of the draft:
(1) draft proposes to use the term "Class-Type" to designate a particular
combination of DS-TE attribute values (which set of BCs apply, which
preemption level applies, which restoration level applies...)
I recommend you use another term to designate these combinations of attributes:
- it would be very confusing to have multiple terms each designating one
"attribute" (preemption, restoration,...) and have one term which designate
both (i) an attribute (sharing the same set of BCs) and (ii) a complete
combination of all attribute values.
- I think we just got to the point where we now have a fine definition of
Class-Type in the DSTE-REQTS document which addresses the concerns
expressed so far (and that we may be able to close on).
So, if you want to designate particular combinations of all the DSTE
attributes, my recommendation would be to leave each attribute with its own
current name (preemption, Class-Types as currently defined in
DSTE-REQTS,...) and give another name to the concept of attribute
combinations ("DSTE Class", "DSTE attribute set", or whatever, but not CT).
Is that acceptable?
Alternatively, if you really like the term CT for your combinations of DSTE
attributes, we could rename the attribute of "sharing the same BCs"
(currently named CT in DSTE-REQTs) into someting like "Constraint Group".
That would be fine for me too. I just want the current DSTE-REQTs concept
to remain, don't care too much about its name.
Would you prefer that?
(2) draft proposes 6 fixed combinations of attributes (ie 6 "DSTE Classes").
I think it would be useful to document these combinations as something that
one SP (or a set of SPs) intend to deploy.
I don't think it would be appropriate to document those as something all
SPs must adhere to.
So my recommendation would be to document those as a BCP.
Is this a reasonable path?
Cheers
Francois