The MPLS WG Archive

["Amir Hermelin" <amir@cwnt.com>]

Feihong,

It's enough to have one reason not to make a certain parameter
mandatory, in order to leave it optional. I'm sure there are other
reasons as well, but here are some:
* unreserved bw:  suppose your domain is only interested in Diffserv-TE
(whatever will come of it), and DS-TE advertises unres bw in different
TLVs. Then, there's no need to prevent turning (BE) unres-bw off.
* max bw and max reservable bw:  same as above.
* te metric:  suppose you don't want to use it (as some implementations
don't).
* color:  if a color hasn't been configured, there might not be a need
to advertise it.
* local and remote ip:  these can be inferred in other ways (although in
IS-IS the situation might be different).

So, in general, leaving them optional is really providing flexibility
(for the future, as well). I do agree, that advertising only LinkID and
LinkType TLVs (or, in the IS-IS case, advertising nothing) is useless to
TE (whether offline, online, or on the moon...).

Hope this helps...

--
Amir Hermelin                  <mailto:amir@cwnt.com>
Charlotte's Web Networks Inc.   <http://www.cwnt.com>


-----Original Message-----
From: Feihong Chen [mailto:fchen6@lucent.com]
Sent: Wednesday, August 22, 2001 7:13 PM
To: dkatz@juniper.net; kireeti@juniper.net
Cc: mpls@UU.NET
Subject: question about the traffic engineering extensions to OSPF


Hi, 

I just realized that the new version of the "Traffic Engineering
Extensions
to OSPF (draft-katz-yeung-ospf-traffic-05.txt)" made some changes on the
Sec2.4.2 (Link TLV). Unlike the previous version of the draft, this
version
proposed that "The Link Type and Link ID sub-TLVs are mandatory, i.e.,
must
appear exactly once. All other sub-TLVs defined here may occur at most
once." (Note, other sub-TLVs includes traffic engineering metric,
local/remote interface ip address, bandwidth information, class/color).

The question that I have here is, if the LSA only advertise the Link
Type
and Link ID as mandatory, then I don't see the capability of OSPF being
a
"traffic engineering helper" because the Traffic Engineering Database
build
from these two information ( which is the worst case ) simply is not
able to
calculate the ERO (i.e. the list of ip addresses), nor the bandwidth for
an
attempted LSP. 

In another word, if the proposed extensions to OSPF is mainly for
traffic
engineering, shouldn't some or all of  "other sub-TLVs" (e.g. bandwidth,
ip
address, color) be necessarily advertised besides the Link Type and Link
ID?
Otherwise, it will end up the traditional OSPF not for TE purpose? 

Maybe I miss something here, looking forward to a clarify to better
understand the purpose of this updated changes.

Thanks a lot,
Feihong Chen