| Hi, Ketan:
The logic is that why we can set router-id equal to 0.0.0.0 to indicate some information in some standards, but we can’t set prefix originator information to 0.0.0.0 to indicate the prefix is unreachable?
Here are again two examples for the usages of router-id’s value as 0.0.0.0 to indicate some information, one is for OSPF another is for IS-IS:
Designated Router ID
The sending router's view of the identity of the Designated Router for this network. The Designated Router is identified by its Router ID. It is set to 0.0.0.0 if there is no Designated Router.
Backup Designated Router ID
The sending router's view of the identity of the Backup Designated Router for this network. The Backup Designated Router is identified by its IP Router ID. It is set to 0.0.0.0 if there is no Backup Designated Router.
If the originating node does not support IPv4, then the reserved value 0.0.0.0 MUST be used in the Router ID field, and the IPv6 TE Router ID sub-TLV [RFC5316] MUST be present in the TLV.
What I insist is that there are already containers that can be used to indicate the unreachable information, why we pursue other non-existing, non-standard container? On Nov 7, 2023, at 18:16, Ketan Talaulikar <ketant.i...@gmail.com> wrote:
Hi Aijun,
I am not sure what "logic" you are looking for while being somewhat dismissive of the arguments/logic provided.
Let us agree to disagree.
At least I've concluded that it is no more fruitful for me to try to convince you. C'est la vie ...
Thanks, Ketan
Hi, Ketan:
There are many examples within IETF that special values has special meanings, please see:
4) LS-Infinity
Then, please state clearly that why we cannot define specific value for prefix originator to indicate the unreachability.
We need the logic that supports your conclusions. Until now, none.
Or anyone else can elaborate the logic more clearly?
Hi Aijun,
I realize that continuing this argument with you is futile.
However, perhaps one last response that I would address not to you but for other WG members (if any one is still following this thread). Hi, Ketan and Les:
There are two sub-TLVs to indicate the source information of the prefix within OSPF——“Prefix Source OSPF Router ID” and “Prefix Source OSPF Router Address”
What’s you refer to is the first sub-TLV, for the second sub-TLV, we haven’t such statements, this is also true for IS-IS, as Les pointed out.
KT> I am not a lawyer. The semantics for Prefix Source OSPF Router Address should be clear to anyone that reads the procedures in Sec 3.
So, contrary to your happiness :) this give us the room to define the specific value to indicate “unreachability”.
And, to Ketan again, until now, you don’t explain clearly that if we can’t define specific value for “unreachability” why can we define the specific value for “LS-Infinity”?
KT> For LS-Infinity - please read RFC2328.
Thanks, Ketan
Ketan –
I am very happy to be wrong in this case.
😊
We are in full agreement.
Les
Hi Les,
I disagree with your reading of RFC9084 (OSPF Prefix Originator).
This document proposes extensions to the OSPF protocol for the inclusion of information associated with the router originating the prefix along with the prefix advertisement. These
extensions do not change the core OSPF route computation functionality.
For intra-area prefix advertisements, the Prefix Source OSPF Router-ID Sub-TLV MUST be
considered invalid and ignored if the OSPF Router ID field is not the same as the Advertising Router field in the containing LSA. Similar validation cannot be reliably performed for inter-area and external prefix advertisements.¶
A received Prefix Source OSPF Router-ID Sub-TLV with the OSPF Router ID field set to 0 MUST be
considered invalid and ignored. Additionally, reception of such sub-TLVs SHOULD be logged
as an error (subject to rate limiting).
As editor of this document, it is absolutely clear to me that we are referring to the sub-TLV and not the prefix advertisement. So, when the value is set to 0, the sub-TLV is invalid and ignored
- the prefix reachability is not at all affected.
This is the reason why I am firmly opposed to using Prefix Originator value 0 for UPA or any other indication.
I hope I am able to convince you :-)
To add to what Ketan has stated…
draft-wang-lsr-prefix-unreachable-annoucement defines the same mechanism for both OSPF and IS-IS i.e., it proposes to use a prefix-originator sub-TLV with address set to 0.0.0.0
to indicate unreachability.
For OSPF, this might be considered compatible with RFC 9084 where it is stated that advertisements with “Router ID field set to 0 MUST be considered invalid and ignored” - though
Ketan has indicated this usage is undesirable.
However, no equivalent statement was ever made for IS-IS in RFC 7794 – so this simply does not work in the case of IS-IS.
I (among others) pointed this out to the authors of draft-wang multiple times over the years, but my feedback was ignored.
Which is an example of why I would like to echo Ketan’s sentiments – let’s please put an end to this non-constructive discussion.
The authors of draft-wang have had many opportunities over the years to respond in a more constructive way to feedback. They were also – as Peter has indicated – given an opportunity
to co-author draft-ietf-lsr-igp-ureach-prefix-announce out of respect for them bringing the problem space to the attention of the WG. They declined – which of course is their right. But they do not have the right to endlessly oppose the consensus of the WG.
Let’s please move on.
Les
Hi Aijun,
As your co-author on the OSPF Prefix Originator RFC, I have stated many times (e.g. [1]) that overloading semantics of Prefix Originator is not a good or clean protocol encoding.
Semantically, it is wrong and a very bad protocol design in my opinion. Going by this logic, tomorrow, someone would want to encode some different meaning for all 1's value in the originator address. We cannot be doing such (IMHO) HACKS in the protocol encodings.
It is better to signal the semantics/meaning explicitly using specific flags that are meaningful.
The authors of draft-ppsenak (now a WG document) agreed to this feedback from WG members and incorporated the U/UP flags in their draft. However, the authors of draft-wang seem
to continue to refuse to accept feedback. It is perhaps one of the reasons why the WG adopted the draft-ppsenak and not draft-wang.
WG chairs, is there a way to put an end to this debate such that it does not continue endlessly?
You castigated Peter for his lack of rigor in his reply to your email, however, I think that was completely unfounded.
Further, your reply to Peter seems to be argument by emphatic assertion, rather than "technical analysis/comparison".
Hi, Peter:
Let’s focus on the technical analysis/comparison for the mentioned issues, and don’t repeat the subjective
comments that without any solid analysis.
Detail replies inline below.
Aijun,
please see inline:
On 06/11/2023 13:23, Aijun Wang wrote:
Hi, all:
Here are some technical questions for the hurry adopted draft about unreachable prefixes announcement:
1) There exists already “prefix originator” sub-TLV to indicate the associated prefix is unreachable,
what’s the advantage of using other undefined, to-be-standardized, to-be-implemented sub-TLV?
many people have already commented on why overloading the “prefix originator” sub-TLV to signal unreachability is a bad idea. Please accept that feedback.
[WAJ] No one gives the technical analysis. Can you explain technically in detail?
You can set the prefix metric to LS-infinity to indicate the unreachability, why can’t we the prefix
originator to NULL to indicate the unreachability?———The key idea for using “prefix originator” is here: if there is no router originate the associated prefix, then it is certainly unreachable. It is more straightforward than the LS_Infinity, and is also more
easily implemented, deployed than the to-be-defined, to-be-standardized sub-TLV.
2) It is unnecessary to define the “UP” flag——if the operator know the unreachable event in advance,
they can also schedule the switchover of the related services in advance. Why bother IGP to transfer such information?
looks like there are folks that see the value in it. I let them to comment more, but I don't necessarily see a problem in an extra bit. If you don't like it, don't use it.
3) There is very limited usage of LS_Infinity in current network. From the operator’s viewpoint,
we will decrease its usage also in future. Then the solution should try their best to avoid their usages——Current solutions instead enhance its usage——It is unacceptable. Let’s keep the network simple.
the reasons for using the LSInfinity for unreachability has been discussed at great length a;ready.
It's the backward compatibility for routers not supporting the new signalling - we need to avoid them interpreting the unreachability as reachability.
[WAJ] My emphasis is that we shouldn’t enhance the usage of LS-Infinity(you propose that the LS-Infinity
metric must be carried) Instead, we should try to fade them out:
1) If all routers within one area/domain all support the new capabilities, we need not require the
summary LSA to carry LS-infinity metric.
2) The Maxage of LSA can also be used to achieve the similar effects of legacy node bypassing.
4) We can’t ignore the partitions scenarios or let’s it go.
if you feel like the partition is the problem, you can write a separate draft and address it there. We are NOT trying to solve it with UPA draft. And for a reason.
[WAJ] They are coupled. If you don’t consider it now, you need to update your proposal later.
[WAJ] I am referring to the balance advertisement of reachable information, as did in the above link,
not the simple statement as the following: “It is also recommended that implementations limit the number of UPA advertisements which can be originated at a given time. “
_______________________________________________
Lsr mailing list
Lsr@ietf.org
https://www.ietf.org/mailman/listinfo/lsr
_______________________________________________Lsr mailing listLsr@ietf.orghttps://www.ietf.org/mailman/listinfo/lsr
|
