Re: [perpass] comments and questions for the group on draft-farrell-perpass-attack-02

[Stephen Farrell <[e-mail]>]

Hi Eliot,

I've trimmed the cc list to perpass. (I'll send a link to your
mail to [e-mail] which is arguably where this discussion
should happen since the draft is in IETF LC. But I guess most
of the folks who care are on here too so its not a huge deal.)

On 12/09/2013 12:29 PM, Eliot Lear wrote:
> These comments follow a thread from the IAB/IESG mailing list.
> 
> There has been some confusion over how this proposed BCP should be taken
> by working groups.  At least one working group chair has inferred, if
> not suggested, that a document before the IESG that does not demonstrate
> it has mitigated pervasive surveillance will get returned.  

Can we establish the reality here? The chair you mean is Mark
Nottingham in this [1] mail to the httpbis list.

   [1] http://lists.w3.org/Archives/Public/ietf-http-wg/2013OctDec/1453.html

I definitely did not read him the way you appear to have and
that distinction matters. If you are the only one to take him
as saying that then I guess you'd agree that your changes
would be based on a fallacy. Maybe Mark can clarify but I think
its already crystal clear that he was not saying "ignore
everything else" - I'd be stunned if that was what he meant.
And I'm sure even if it were, the httpbis WG participants
would either ignore that or beat him up for it (which they've
not done).

That said, it is clear that individual WGs and other IETF
activities will have to figure out what it means to properly
consider pervasive monitoring as an attack, but the draft says
that already I think. What I read Mark as saying was that if
the httpbis WG ignored or didn't properly consider pervasive
monitoring and if this BCP gets approved then that could
impact on their schedule. There's nothing surprising there at
all IMO.

> That
> wouldn't be appropriate, in my opinion.  A working group must consider
> the operational realities that it is attempting to code to.  There are
> many tensions to consider, not just a narrow view of network management
> (a phrase that might itself be left to ambiguous interpretation).  

"Operational realities" is far more vague IMO. Perhaps even dangerously
so since it could be used to attempt to justify pretty much anything.
And as Eric Burger already noted, what you propose would also seem to
fly in the face of RFC 2804, which is entirely antithetical to the
intent of this draft.

I think the current wording is much better than your proposed changes.

> I
> have in mind as an example a service provider that does
> transparent/intercepting HTTP caching for purposes of bandwidth
> preservation in bandwidth-constrained environments.  

The httpbis WG are discussing how to handle proxies. That's a
complex and involved thread and I don't think any of us will
benefit from re-running the discussion here. Folks who care
about that should subscribe to that list and discuss it there.

> Another example
> would be an environment in which security and governance are key.

I've no idea what that means. If you mean something that's
broader than http, then what? If you just mean for http then
again that's better dealt with in httpbis.

> 
> As such, I propose the following two changes to clarify the text in
> Section 2:
> 
> 2nd para:
> 
> OLD:
>>    This BCP simply records the consensus to design
>>    protocols so as to mitigate the attack, where possible.
> 
> NEW:
> 
>>    This BCP simply records the consensus to design
>>    protocols so as to mitigate the attack, taking into account
>> operational realities of network operators.
> 
> Next para:
> 
> OLD:
>>    More limited-scope monitoring to assist with network management that
>>    is required in order to operate the network or an application is not
>>    considered pervasive monitoring.  There is though a clear potential
>>    for such limited monitoring mechanisms to be abused as part of
>>    pervasive monitoring, so this tension needs careful consideration in
>>    protocol design.  Making networks unmanageable in order to mitigate
>>    pervasive monitoring would not be an acceptable outcome.  But
>>    equally, ignoring pervasive monitoring in designing network
>>    management mechanisms would go against the consensus documented in
>>    this BCP.  An appropriate balance will likely emerge over time as
>>    real instances of this tension are considered.
> 
> NEW:
> 
>>    More limited-scope monitoring or other services or to assist with
>> network operations that
>>    is required in order to operate the network or an application is not
>>    considered pervasive monitoring,
>>    There is though a clear potential
>>    for such mechanisms to be abused as part of
>>    pervasive monitoring, so this tension needs careful consideration in
>>    protocol design.  Making networks unmanageable in order to mitigate
>>    pervasive monitoring would not be an acceptable outcome.  But
>>    equally, ignoring pervasive monitoring in designing network
>>    management mechanisms would go against the consensus documented in
>>    this BCP.  An appropriate balance will likely emerge over time as
>>    real instances of this tension are considered.

To be clear, I'd prefer we make neither change.

> And so someone operating a small ISP with a cache in Madagascar or the
> SP trying to manage bandwidth on a train can also have their
> considerations taken into account.  It's for a working group to decide
> how far to accommodate them, of course.

And that's already clearly the case. This proposed BCP doesn't
do away with either working groups or common sense.

S.

> 
> Eliot
> 
> _______________________________________________
> perpass mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/perpass
> 
>