[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
RE: [PWE3] Inter-op is not necessary
Folks,
It looks to me we went through this type of situation
in the past in more than one instance and I believe that
without a clear analysis, guidelines/process and decision
at whole ietf level (not each time leaving *just* the wg to
make the choice -though nothing wrong with that if it succeeds
in making such choice in a timely manner!) I am afraid we will again
encounter such situations and the same arguments will be
played again and again.
>From past experience, at certain point it is not really the
technical arguments that count in this debate but really which
factors are deemed important for which group supporting a
given proposal (including the group that looks at just
complexity of standardization efforts).
For example these factors can be:
- let the market decides (though "market decides" needs
some form of definition).
- avoid wg extra efforts (like define interoperability
mechanism, avoid duplicate efforts, etc.
- try to always build on existing close standards in incremental way
(note that if this argument was really strong CR-LDP will have
made it ;-)).
- adopt solutions that have significant market preference
(again market may prefer several ones) and market
preference can shift with time.
- etc...
It is really which factors we (ietf in general and pwe3 in
particular) think is *important* that will guide the final choice.
My suggestion is instead of rehashing the same previous debates,
why not have ietf/iesg, both? look at this problem
and provide some clear guidelines/process (or decisions)
and let pwe3 members work on what they believe are interested in
(one or two proposal, etc) in early stages.
Maybe one completely different outcome of such analysis
is to acknowledge that these situations will happen in certain
situations at early stages of standard efforts and it maybe
difficult to make clear cut choice, then document
the factors that wg used to adopt
which proposal and revisit such decision within a time
period, or any other proposal...
Hamid.
> -----Original Message-----
> From: pwe3-bounces at ietf.org [mailto:pwe3-bounces at ietf.org] On
> Behalf Of Gray, Eric
> Sent: Tuesday, November 15, 2005 6:35 PM
> To: 'Andrew G. Malis'; LE ROUX Jean-Louis RD-CORE-LAN
> Cc: Yaakov Stein; pwe3 at ietf.org; Mark Townsley
> Subject: RE: [PWE3] Inter-op is not necessary
>
>
> Andy,
>
> There is a bit of anachronism in this interpretation. As
> I recall from the time, it might be slighly more accurate to
> say the most credible router vendors went the same way - but
> that may seem to cast aspersions against other router vendors
> in the arena at the time - which would be at least unfair. It
> might also seem to imply that credibility was derived from
> technical correctness and that would be a bit of a stretch.
>
> When the majority of wrangling over protocol choices for
> traffic engineering was taking place, one of the two router
> vendors that chose to go the "same way" was a start-up
> company just having obtained their second round of funding
> for - I seem to recall a total investment of about 51 million
> dollars (hardly enough to establish them as a dominant router
> vendor - no matter what occurred later).
>
> What made that vendor as credible as any other was the
> clear alignment they had with one of the more credible
> service providers (which also happened to be the one
> fostering the idea of traffic engineering in the first
> place). Other router vendors were less
> "credible" because they were not getting traction with service
> providers (especially the one that was the largest customer that
> had already indicated an intent to pay real money for TE-capable
> routing equipment).
>
> In retrospect, the router vendors that chose to go with
> the RSVP-TE signaling choice ended up as the dominant router vendors
> - and that (I believe) is the result of having the market
> make the choice.
>
> I think we all know that the IETF leadership is under some
> pressure to "make a choice". However, unless it is really
> clear that a choice can be made at this time, the IETF may
> not be the right place to make that choice. As much as
> people like to make the "two protocols is worse than none"
> argument and use CR-LDP and RSVP-TE as an example, any number
> of things could have made things come out differently in that case.
>
> --
> Eric
>
>
> --> -----Original Message-----
> --> From: Andrew G. Malis [mailto:andymalis at comcast.net]
> --> Sent: Tuesday, November 15, 2005 5:50 PM
> --> To: LE ROUX Jean-Louis RD-CORE-LAN
> --> Cc: Andrew G. Malis; Gray, Eric; Mark Townsley; Yaakov
> --> Stein; pwe3 at ietf.org
> --> Subject: RE: [PWE3] Inter-op is not necessary
> -->
> --> JL,
> -->
> --> Actually, some vendors did do both before it was clear
> --> where the market was going - and in this case, we had the
> --> advantage that both of the dominant router vendors went the
> --> same way, as did their customers. But that was really the
> --> exception rather than the rule.
> -->
> --> Cheers,
> --> Andy
> -->
> --> --------
> -->
> --> At 11/15/2005 22:32 +0100, LE ROUX Jean-Louis RD-CORE-LAN wrote:
> --> >Hi Andrew,
> --> >
> --> > >I'm not a fan
> --> > > of "let the market decide"
> --> > > because in the end, as a vendor you have to do both -
> --> >
> --> >Look at the RSVP-TE vs CR-LDP case, this is a good example, a bit
> --> >related to our case here by the way:-), were the market
> --> decided, and it
> --> >seems to me that not many vendors implemented both...
> --> >
> --> >Regards,
> --> >
> --> >JL
> --> >
> --> >
> --> > > -----Message d'origine-----
> --> > > De : pwe3-bounces at ietf.org
> --> [mailto:pwe3-bounces at ietf.org] De la part
> --> > > de Andrew G. Malis Envoyé : lundi 14 novembre 2005
> --> 21:59 À : Gray,
> --> > > Eric Cc : Mark Townsley; 'Yaakov Stein'; pwe3 at ietf.org
> --> Objet : RE:
> --> > > [PWE3] Inter-op is not necessary
> --> > >
> --> > > Eric,
> --> > >
> --> > > Yeah, but we don't want to restrict it to a single
> --> administrative
> --> > > domain. I also agree with many of the others in the
> --> meeting - let's
> --> > > choose one and move on. I'm not a fan of "let the
> --> market decide"
> --> > > because in the end, as a vendor you have to do both -
> --> especially if
> --> > > big vendor a goes one way, big vendor b goes the other,
> --> and everyone
> --> > > else needs to interoperate with both a and b. And most
> --> times, even
> --> > > the big vendors find that their customers want them to
> --> interoperate
> --> > > with their competition, so they end up doing it both
> --> ways as well.
> --> > >
> --> > > The only way we should approve both is if we can define
> --> a completely
> --> > > dynamic interface between the two that requires no more
> --> provisioning
> --> > > by the service providers than if we had only chosen
> --> one. Otherwise,
> --> > > we're shortchanging the operators and the rest of the
> community.
> --> > >
> --> > > Cheers,
> --> > > Andy
> --> > >
> --> > > -------
> --> > >
> --> > > At 11/14/2005 18:39 -0200, Gray, Eric wrote:
> --> > > >Yaakov,
> --> > > >
> --> > > > This is a very narrow view of the purpose for
> --> > > >multi-segment pseudo-wires. There are applications
> --> for pseudo-wire
> --> > > stitching within
> --> > > >a single administrative domain.
> --> > > >
> --> > > >--
> --> > > >Eric
> --> > > >
> --> > > >--> -----Original Message-----
> --> > > >--> From: pwe3-bounces at ietf.org
> --> [mailto:pwe3-bounces at ietf.org] On
> --> > > >--> Behalf Of Yaakov Stein
> --> > > >--> Sent: Monday, November 14, 2005 3:14 PM
> --> > > >--> To: Mark Townsley; pwe3 at ietf.org
> --> > > >--> Subject: RE: [PWE3] Inter-op is not necessary
> --> > > >-->
> --> > > >-->
> --> > > >--> > To be clear, I said that "coexistance" was a
> --> necessity to be
> --> > > >--> documented by the WG.
> --> > > >--> > Whether coexistance equates to interworking
> --> depends on the MS
> --> > > >--> requirements.
> --> > > >-->
> --> > > >--> Since the whole idea of MPLS MS-PWs is to facilitate the
> --> > > >--> interworking of different operators, as far as I see
> --> > > >--> coexistence must mean either
> --> > > >-->
> --> > > >--> 1) fully interwork
> --> > > >--> 2) mandate that they never have to cooperate.
> --> > > >-->
> --> > > >--> The latter would probably end up with everyone having to
> --> > > >--> support both signaling protocols, and often setting up
> --> > > additional PWs just
> --> > > >--> to circumvent the limitation.
> --> > > >-->
> --> > > >--> Y(J)S
> --> > >
> --> > >
> --> > > _______________________________________________
> --> > > pwe3 mailing list
> --> > > pwe3 at ietf.org https://www1.ietf.org/mailman/listinfo/pwe3
> --> > >
> -->
>
> _______________________________________________
> pwe3 mailing list
> pwe3 at ietf.org
> https://www1.ietf.org/mailman/listinfo/pwe3
>
>
_______________________________________________
pwe3 mailing list
pwe3 at ietf.org
https://www1.ietf.org/mailman/listinfo/pwe3