On Jun 23, 2011, at 6:27 PM, Randy Bush wrote:
>> Section 2 says a serial number is:
>> A 32-bit monotonically increasing ordinal which wraps
>> from 2^32-1 to 0.
>> Section 5.10, it says:
>> An implementation which uses a fine granularity of time for the
>> Serial Number might never change the Cache Nonce.
>> Which is correct? The latter is not a monotonically increasing ordinal.
>
> both are. and this is a problem because?
...because the spec is inconsistent. If a router is expecting the first and the
cache is doing the second, the router would decide the cache is noncompliant.
> the nonce is not specified to be increasing or decreasing. it is a
> nonce, not a counter.
Wearing my security pedant hat, it is not a nonce at all. It is a
randomly-selected value that is used repeatedly as a selector. We agree that it
is not a counter. It is not useful to change all the uses of "nonce" in this
spec to something more correct.
>> Section 4 says:
>> When the cache updates its database, it sends a Notify message to
>> every currently connected router.
>> Section 6.2 says:
>> The cache server SHOULD send a notify PDU with its current serial
>> number when the cache's serial changes
>> The first is mandatory, the second is advisory; they should match.
>
> there is no MUST in the *overview* in sec 4. notify is not mandatory.
Good. Then section 4 should say "database, it should send ..."
>
>> In section 5.3, there seems to be an implicit "MUST flush all data",
>> like the explicit one in section 5.10. It should probably be explicit
>> here as well or, if not, the document should say why a router might
>> not flush when getting data after a reset.
>
> from the currently published version -12
>
> 5.3. Reset Query
>
> Reset Query: The router tells the cache that it wants to receive the
> total active, current, non-withdrawn, database. The cache responds
> with a Cache Response PDU (Section 5.4).
>
> i see nothing about flushing anything. restart session, please.
This confuses me. What does "restart session" mean? Section 5.10 says:
If, at
any time, either the router or the cache finds the value of the
nonces they hold disagree, they MUST completely drop the session
and the router MUST flush all data learned from that cache.
That sounds like "restart session" implies "MUST flush". Being a bit more
explicit in 5.3 and 5.10 would help implementations interoperate.
>> Section 5.10 says:
>> If, at
>> any time, either the router or the cache finds the value of the
>> nonces they hold disagree, they MUST completely drop the session
>> and the router MUST flush all data learned from that cache.
>> Why does the session need to be dropped?
>
> because it is not the same session. e.g. serials are not commensurate.
OK, maybe I'm confused by what you mean by "session", and it doesn't appear to
be to defined in the protocol.
>> This could cause delay when re-authenticating.
>
> if the nonce was off i sure would want to re-auth
If you are using transport that assures integrity, there is no reason to
believe that you are no longer talking to the same party just because one of
you forgets the nonce or otherwise gets out of sync. What attack are you trying
to prevent by doing a re-authorization?
>> A non-trivial editorial issue: the draft uses "commensurate" in many
>> places that does not match any of the definitions I can find in my
>> dictionaries. I *think* that the draft means "the same", but that is
>> not clear.
>
> no. it means drawn from the same number space. i.e. if the nonce
> differs, the cache and the router mean very different things by serial
> value 42.
Right, but that's not what "commensurate" means. The common definition across
multiple dictionaries is "proportionate" or "of similar size", and that is
definitely *not* what you mean here. I don't even find your definition in my
math dictionary, and the math term "commensurable" is not applicable here.
--Paul Hoffman
_______________________________________________
sidr mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/sidr