[Sidrops] Re: Local AS in ASPA validation (draft-ietf-sidrops-aspa-verification-19)

["Borchert, Oliver (Fed)" <[e-mail]>]

To avoid multiple different emails, I answered here the questions you had in the previous email.


  *   Can you explain why it would be unreasonable to suggest a limit of
somewhere between 4,000 and 10,000 - describing clearly that some kind
of limit (A) must exist and (B) be periodically reviewed.

I am not opposing a suggestion as it is currently made in the profile draft. Said that, I strongly believe it is something to be handled by the RP before we even talk about 8210bis.


  *   I'm quite 0concerned about CAs (accidentally) DoSing validators and/or
routers with overly long Provider listings.

Define accidentally. Do you mean a BUG? Or am I missing something?


  *   Do you have some other approach in mind that avoids suggesting limits?

There are ways to propose a dynamic “limit” that is future-safe. I normally did not want to go down this route, but here you go:

Modify the Reset Query PDU (2) my using the zero field (2 bytes) as the max length indicator for ASPA provider lists.

0          8          16         24        31
.-------------------------------------------.
| Protocol |   PDU    |                     |
| Version  |   Type   |     Max Providers   |
|    2     |    2     |    ( 0 = no limit)  |
+-------------------------------------------+
|                                           |
|                 Length=8                  |
|                                           |
`-------------------------------------------'

This signals the maximum number of ASPA provider ASes the router can currently process. 0 = no limitation.

Now, if the RP generates a list of more than “Max Providers,” the RP will send a non-fatal Error PDU to the router stating which Customer-AS has more than “Max Providers” providers in the list and the number of providers. The ASPA PDU will be held back and not transmitted to the router.
Both parties SHOULD create a log entry on their side.

This is one way to be future-proof without the need to change 8210bis in the future in case a revision of the limit results in an increase. It also allows for backward compatibility.


Said all that, it still does not answer what it means if an RP cannot send the valid full list once it exceeds the limit. At least the error does provide the router with a log-able event which can be addressed by an operator.



  *   ps. Autonomous System Numbers can be more than 7 digits, right?
https://github.com/usnistgov/NIST-BGP-SRx/blob/e978750b7be3b51a2042517dd6c0568f56e4e03c/srx-server/src/server/aspa_trie.c#L444

This indeed is a BUG and will be fixed, thanks for catching it,

Oliver


From: Job Snijders <[e-mail]>
Date: Thursday, December 5, 2024 at 12:24 PM
To: Sriram, Kotikalapudi (Fed) <[e-mail]>
Cc: Claudio Jeker <[e-mail]>, Borchert, Oliver (Fed) <[e-mail]>, Maria Matejka <[e-mail]>, [e-mail] <[e-mail]>
Subject: Re: [Sidrops] Re: Local AS in ASPA validation (draft-ietf-sidrops-aspa-verification-19)
On Thu, 5 Dec 2024 at 18:21, Sriram, Kotikalapudi (Fed) <[e-mail]<mailto:[e-mail]>> wrote:
Claudio,

>Not having a limit in the draft will result in serious deployment issues.
>So we better sit down define a limit and make sure it is large enough for any sensible case. The one ISP that has every other ISP as provider need to justify why. >The current numbers of 4000 - 10000 are more than a magnitude higher than what is currently seen in the internet.

Oliver has already said: "The proper way going forward is to modify section 6 of the ASPA Profile draft to require that an upper boundary MUST NOT exceeded 10,000 provider ASes. (Or 4,000 if this is more reasonable). This can easily be updated later to a higher number."

If a required limit is specified in the ASPA profile draft, is there a need to specify it again in 8210-bis?


Please see
https://mailarchive.ietf.org/arch/msg/sidrops/loHK1tvGHZuwFIvDkmUfZx_zid4/

Multiple ASPA objects can contribute to a single PDU, and there also is SLURM to consider.

Kind regards,

Job