Hi Anton,
On 10/23/13 5:09 AM, "Anton Smirnov" <[email protected]> wrote: > Hi Acce, > you are just not formulating your objections in the correct way. For >example, if specification said that Tag TLV must not exceed 64 Kb in >size that would technically put a bound but this would be both pointless >and wouldn't satisfy you. The applicability statement in section 3 of RFC 4970 is clear. It has nothing to do with how I formulate my objections ;^) > > Another part is why RI LSA is being singled out - why unbound data >in, say, Network LSA is OK and unbound data in RI LSA is not? Especially >given that RI LSA can be extended to the adjacent LSIDs and the Network >LSA cannot. This is a terrible analogy. The Network LSA has a single purpose and is bounded by the number of OSPF routers on a LAN. In this case, the TLV has no bound and implementors could come up with all sorts of clever ways to encode the data. > > Tag data ARE expected to be very small and very stable, so the >choice of RI LSA to advertise them is very reasonable. I've heard requirements that contradict this statement in this E-mail thread. Thanks, Acee > >Anton > > >On 10/22/2013 10:05 PM, Acee Lindem wrote: >> I don't disagree that the typical use case is a single tag with the >> likelihood of mult-tag use cases diminishing exponentially as the >> number of tags increases. My point is that unbounded TLVs MUST NOT be >> included in the OSPF RI LSA. What part of that is hard to >> understand? I think that 16 is a reasonable maximum and that beyond >> 16 would imply encoding ulterior information that should have its own >> TLV or LSA anyway. Acee >> >> On Oct 22, 2013, at 3:48 PM, Anton Smirnov wrote: >> >>> Hi Acee, it looks to me that the most probable deployment will use >>> 1 tag. Router advertising 100 tags already sounds unreasonable. >>> Defining new LSID to originate LSA with (typically) only 4 bytes of >>> useful information is not optimal. Choice of RI LSA to advertise >>> some small data is reasonable. RI LSA is far from getting too big. >>> If there is a concern of RI LSA overfilling then we can extend >>> range of opaque IDs - but is it really necessary at this point? >>> >>> Anton >>> >>> >>> On 10/21/2013 09:55 PM, Acee Lindem wrote: >>>> I think we are in a circular argument here and I'm not discuss >>>> this independently with each of the authors. Either you have to >>>> limit the number of tags, define a new LSA, or do the work to >>>> make RI LSA multi-instance. All are viable alternatives with >>>> differing pros and cons - including it in the existing RI LSA is >>>> not a viable alternative. Remember to request a session if you >>>> plan to present it at IETF 88. >>>> >>>> Thanks, Acee > _______________________________________________ OSPF mailing list [email protected] https://www.ietf.org/mailman/listinfo/ospf
