Sandra Murphy wrote:
> The relative frequency of use of AS_SETs is interesting, but not really
> germane to the point here.
>
> If we were trying to develop a protection for AS_SETs then we might want
> to ask the engineering question of where and how often they were used.
>
> But for the purpose of validating received updates, we need a rule for
> what is done with AS_SETs that appear in the AS_PATH origin. Lack of
> rules leaves opportunities for deliberate or accidental mischief.
>
> AS_SETs might not be used very often, but that doesn't stop someone from
> using AS_SETs deliberately with malicious intent.
right so as a starting point:
"AS_SET in an origin is unvalidatable."
how about that? (I think this is fine since:
1) they aren't used in production very much anymore
2) where used, they seem to be mis-used
3) the rules for how you do verification/validation of an AS_SET are at
best murky.
-chris
(regular user)
> --Sandy, making a technical point as wg member and a document
> completeness point as wg chair.
>
> On Wed, 28 Apr 2010, Warren Kumari wrote:
>
>>
>> On Apr 28, 2010, at 2:37 PM, Sandra Murphy wrote:
>>
>>>
>>>
>>> On Thu, 29 Apr 2010, Geoff Huston wrote:
>>>
>>>> Thank you for this response. As I had noted earlier, if you had made
>>>> it clearer in which role you were posting in these discussions it
>>>> would be easier for others, or at least myself, to understand when
>>>> you were making pronouncements as WG chair and when you were asking
>>>> questions and airing opinions as an individual participant.
>>>>
>>>> If proxy aggregation is "not part of our work" then the outcome of
>>>> such actions, namely AS Sets in the AS_PATH is likewise "not part of
>>>> our work." I will edit the roa validation draft accordingly.
>>>
>>> Actually, no, that is not the case.
>>>
>>> When I say that proxy aggregation is not part of our work, I mean
>>> that sidr does not promise to protect BGP announcements formed by
>>> proxy aggregation.
>>>
>>> While it is true that AS_SETS are principally generated in BGP
>>> updates by an AS that is doing proxy aggregation, AS_SETS are still
>>> part of the BGP protocol and might appear in a received AS_PATH.
>>>
>>> Until and unless the idr working group decides to eliminate that
>>> feature from the BGP protocol, the sidr route validation must still
>>> make a statement about what happens when the AS_PATH origin is an
>>> AS_SET.
>>>
>>> That is necessary for completeness. Furthermore, without such a
>>> statement, implementers could make different decisions as to how to
>>> handle such cases and attack paths could result.
>>>
>>> The sidr wg is not now considering anything but origin authorization,
>>> so an AS_PATH segment that is an AS_SET but is not the origin is not
>>> of our concern at all.
>>
>> Just for giggles I decided to see how how widely AS Sets were being
>> used by having a quick look at route views data (yes, which I realize
>> is incomplete, etc).
>>
>> I used an as-path regexp to only get AS paths that included an AS_SET
>> and then did some grep, sed and awk hackery to get a handle on the data.
>>
>>
>> I see zero AS_PATHs where the AS_SET is not the origin (i.e, where the
>> AS_SET segment is not followed by i (internal), e (external) or ?
>> (unknown):
>>
>> wkum...@colon:~/tmp$ cat assets | grep '} [^ie\?]'
>>
>> I see 39 unique paths where the "AS_SET" is only a single AS:
>> wkum...@colon:~/tmp$ cat assets | egrep "^\*" | sed
>> 's/.*{\(.*\)}.*/\1/' | sort | uniq | grep -v ',' | wc -l
>> 39
>>
>> and 21 unique paths where the AS_SET contains multiple ASs:
>> wkum...@colon:~/tmp$ cat assets | egrep "^\*" | sed
>> 's/.*{\(.*\)}.*/\1/' | sort | uniq | grep ','
>> 11060,12262
>> 1239,17079,27773
>> 17746,64512
>> 209,25729
>> 21135,49476
>> 25019,41426
>> 25912,65427
>> 31812,65473
>> 33363,65129,65188
>> 3356,8928,22351
>> 3491,15412,19710
>> 35821,35821,35821,35821
>> 45514,45514,45514
>> 45514,45514,45514,65081
>> 4558,30994,33770
>> 46124,46136
>> 50693,64601,64604,64607,64608,64609,64610,65505
>> 64520,64521
>> 7470,24128,38865,45199
>> 8002,15412
>> 8508,24748,35434
>>
>> Of these, only:
>> 11060,12262
>> 1239,17079,27773
>> 209,25729
>> 21135,49476
>> 25019,41426
>> 3356,8928,22351
>> 3491,15412,19710
>> 35821,35821,35821,35821
>> 45514,45514,45514
>> 4558,30994,33770
>> 46124,46136
>> 7470,24128,38865,45199
>> 8002,15412
>> 8508,24748,35434
>>
>> do not contain private AS numbers. Some of the above contain only a
>> single AS repeated multiple times (like 35821).
>>
>> I don't want this to devolve into a discussion on the merits of
>> aggregation, if AS_SETs should be deprecated or anything similar, this
>> was just because I wanted know for my own sake what the numbers were
>> and figured others might be interested too....
>>
>> W
>>
>>
>>
>>>
>>>
>>>
>>> --Sandy
>>>
>>>
>>>
>>>>
>>>>
>>>> On 29/04/2010, at 3:16 AM, Sandra Murphy wrote:
>>>>
>>>>> I have said that it was the consensus even before sidr was
>>>>> officially a wg that proxy aggregation would not be part of our
>>>>> work. And I pointed to the email archive for that discussion.
>>>>>
>>>>> You have said that you do not believe that it is possible at this
>>>>> point, so I don't think that you are unhappy with that statement.
>>>>>
>>>>> So I am confused as to what further statement I could make that
>>>>> would help at this point.
>>>>>
>>>>> Can you point me to where you still see a problem?
>>>>>
>>>>> --Sandy
>>>>>
>>>>> On Wed, 28 Apr 2010, Geoff Huston wrote:
>>>>>
>>>>>> three weeks ago I asked:
>>>>>>
>>>>>>>
>>>>>>> It seems to me that the essential requirements for securing proxy
>>>>>>> aggregation are missing at this stage, which makes it somewhat
>>>>>>> difficult for SIDR to work on mechanisms without some re-spinning
>>>>>>> of the SIDR WG Charter (or some other WG) that would permit the
>>>>>>> preliminary work on security requirements relating to proxy
>>>>>>> aggregation to come first.
>>>>>>>
>>>>>>> So my question to the WG Co-chairs is: is work on securing Proxy
>>>>>>> Aggregation within the current SIDR charter? If so, on what basis?
>>>>>>
>>>>>> I would hope that by now the WGchairs have had sufficient time to
>>>>>> consider this question, so I'd like to ask once more: Is work on
>>>>>> securing Proxy Aggregation within the current SIDR Charter? If
>>>>>> so, on what basis?
>>>>>>
>>>>>>
>>>>>> regards,
>>>>>>
>>>>>> Geoff
>>>>>>
>>>>>>
>>>>
>>>>
>>> _______________________________________________
>>> sidr mailing list
>>> [email protected]
>>> https://www.ietf.org/mailman/listinfo/sidr
>>
>> --
>> Eagles soar but a weasel will never get sucked into a jet engine
>>
>>
_______________________________________________
sidr mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/sidr