Re: 6PE-Alt

["Vishwas Manral" <vishwas.ietf@gmail.com>]

Hi Francois,

Thanks a lot for the reply.

It seems the only thing the 6PE draft does is that it allows different
label allocation policies (for different routes), when no label
allocation is required at all.

If you think people will still use the 6PE approach even when they can
achieve the same without any new signaling extensions (a new AFI/SAFI
pair) then I am fine to position the draft as an option to your draft.
Do let me know if you feel that way?

Thanks,
Vishwas

On Jan 31, 2008 11:00 AM, Francois Le Faucheur IMAP <flefauch@cisco.com> wrote:
> Hi Vishwas,
>
> I think I've brought up my main points in the previous message so
> will leave it at that for now.
>
> On 31 Jan 2008, at 18:29, Vishwas Manral wrote:
>
> > Hi Francois,
> >
> > As the 6PE-Alt does not require any additional functionality from
> > BGP-IPv6 (yes that is the signaling required - but that is the same
> > required for normal BGP IPv6. For 6PE we use a new AFI/ SAFI and send
> > labels uselessly. So please do not say that the claim no signaling
> > changes are required for 6PE-Alt  is a stretch), and that is why we
> > have inter-operable implementations of 6PE-Alt too. It is not about
> > interoperable implementations alone though. I however am surprised how
> > this simple mechanism was not captured earlier in the review or coding
> > process.
> >
> > We do not need to transport labels at all which are now being
> > transported. Also as the peer decides the label I use, we may have to
> > configure different labels from different peers (thus using bigger
> > keys/ more memory in the tables and a lot of other overload). There is
> > a lot of overhead of signaling and dataplane without any necessity.
> >
> > I also have issues with the 6VPE RFC too but let us tackle issues one
> > at a time. Also please note 6VPE has the concept of VRF's and is used
> > to give VPN services, while VPE is used for basic IPv6 connectivity
> > over MPLS IPv4 coulds. So please do not confuse the 2.
>
> I didn't really feel confused about the 2, but thanks for the advise ^)
>
> Cheers
>
> Francois
>
>
> >
> > Thanks,
> > Vishwas
> >
> > On Jan 31, 2008 8:53 AM, Francois Le Faucheur IMAP
> > <flefauch@cisco.com> wrote:
> >>
> >> On 31 Jan 2008, at 15:25, Vishwas Manral wrote:
> >>
> >>> Hi Francois,
> >>>
> >>> The document I have written does not use signaling
> >>
> >> It is a bit of a stretch to claim that 6PE-Alt does not use
> >> signaling.
> >> 6PE-Alt and 6PE both use MP-BGP to signal IPv6 reachability.  6PE-Alt
> >> would simply save you one field (the label value) inside the same
> >> advertisement.
> >>
> >> On the other hand, 6PE-Alt would restrict operations to only one of
> >> the options allowed in 6PE. The reasons 6PE allowed multiple label
> >> allocation is precisely because different PEs may want to use
> >> different label allocation policies.  For example, one 6PE router may
> >> prefer to allocate per-prefix label in order to be able to forward
> >> packets via label switching as opposed to performing IPv6 lookup. 6PE
> >> allows each PR router to use whatever label allocation scheme it sees
> >> best.
> >>
> >>
> >>> and does not define
> >>> a new AFI/SAFI functionality to achieve the 6PE.
> >>
> >> It is a bit of a stretch to imply that 6PE defines a new AFI/SAFI.
> >> 6PE uses the IPv6 AFI and the SAFI defined in RFC3107 (which for
> >> memory dates back from 2001) precisely for the purpose of binding an
> >> MPLS label when advertising IPv6 reachability.
> >>
> >>>
> >>> The difference is obvious, we get the same functionality
> >>
> >> my impression is that you get a restricted subset of the
> >> functionality (ie forcing every router to use that single label
> >> allocation scheme). This had been discussed before and decided that
> >> such a restriction was undesirable and unnecessary.
> >>
> >>> without
> >>> adding any new signaling mechanism.
> >>
> >> my impression is that both use the same MP-BGP signaling mechanism
> >> except one would save a few bytes in the same MP-BGP advertisements.
> >>
> >> There are multiple interoperable implementations of 6PE today and am
> >> not aware of significant issues with the current 6PE control plane.
> >>
> >> Cheers
> >>
> >> Francois
> >>
>
>