RE: What we disagree on RE: TE Requirements Draft-ELSP

[Shai Herzog <sherzog@hyperchip.com>]

Neil,

I would second your proposal that TE and VPN work would be tightly coupled
(as VPNs are really just a form of TE, externally facing customers).

I would also agree that I see the usefullness of L-LSPs as well as E-LSPs
with single OA. 

In fact, I would rather drop the E/L LSP name completely and instead use
only one name: DS-LSP. Since any DS LSP needs a mapping between 8 EXP bits
and 64 DSCP space, we can either define a global mapping (pre-configured
identically for all LSR in the domain), or signal it at LSP creation (a la
L-LSP). 

So, we'll have only a single type of DS-LSP, with global pre-configured
setting, and per-LSP signalled override. Simple to implement and effective.

BTW: we haven't discussed it really, but I believe Best Effort (BE) is
included in any OA and can be considered as part of the OA (BE+AF1) would be
considered a single OA. I can justify it by saying that BE has no
constraints, so it will never "take away" or "conflict" with any other QoS
mechanism. 

Does anyone see a problem with that?

shai

>-----Original Message-----
>From: neil.2.harrison@bt.com [mailto:neil.2.harrison@bt.com]
>Sent: December 6, 2001 6:09 AM
>To: Shai Herzog; flefauch@europe.cisco.com; te-wg@ops.ietf.org
>Subject: RE: What we disagree on RE: TE Requirements Draft-ELSP
>
>
>The key thing that is missing from this debate is the 
>relationship between:
>-	public services
>-	private services, ie VPNs
>
>SPs would like to integrate their 'Internet' and 'VPN' networks but
>currently this is proving difficult....so IMO all future work 
>in this area
>ought to recognise/address this issue.
>
>I can see a good solution for the former by aggregation in internal
>public-VPNs using L-LSPs (per PHB).
>
>I can't see any good solution for private VPNs when LDP is 
>used (or indeed
>unification of private/public VPNs)....I have explained 
>several times why
>this is true, so I won't bore you all again.
>
>I can see a potential good solution for private VPN based on 
>E-LSPs....and
>again I have provided the reasons on previous mails....and its 
>seems this
>could be the unifying solution too.
>
>regards, Neil
>
>> -----Original Message-----
>> From: Shai Herzog [mailto:sherzog@hyperchip.com]
>> Sent: 04 December 2001 18:31
>> To: 'Francois Le Faucheur'; 'te-wg@ops.ietf.org'
>> Subject: Re: What we disagree on RE: TE Requirements Draft-ELSP
>> 
>> 
>> At 1010 AM 04/12/2001, Francois Le Faucheur wrote
>> >Nabil and Shahram,
>> >
>> >The one option we disagree on is
>> >(iiia) Using E-LSPs with traffic from multiple OAs with 
>> multiple BW and 
>> >single value for all other attributes (preemption, CT, affinity...).
>> 
>> I for one disagree that E-LSPs with Multiple OAs and single BW are
>> technically (or theoretically) sound. Not to use the cliche of apples
>> and oranges, pretty much it is equivalent to:
>> 
>> Given a box, how many watermelons and cherries can you fit inside it?
>> 
>> Since different OAs are disjoint, and draw from different pools,
>> how do you propose we split the single BW parameters among 
>> the multiple
>> OAs?
>> 
>> Shai
>> 
>