Speaking as a coauthor of the Group B document, it's going to get
extremely confusing if we get individual comments in a thread that
started out referring to Group A. Our drafts should appear in the
I-D archive in the next couple of days; it was sent in on Friday.
At 11:43 AM -0500 3/4/02, Lloyd Wood wrote:
>On Mon, 4 Mar 2002, Tom Scott wrote:
>
>> was typo. this url works:
>> http://partner.unispherenetworks.com/rrg/
>>
>> Regarding Group B, I'd like to draw attention to section 10.9:
>>
>> Is it possible to apply a control theory framework, and analyze the
>> stability of the control system of the whole network domain, for e.g.
>> convergence speed and the frequency response, and then use the
>> results from that analysis to set the timers and other protocol
>> parameters.
>>
>> Control theory could also play a part is QoS Routing, by modifying
>> current link state protocols with link costs dependent on load.
>> Control theory is used to increase the stability of such systems.
>>
>> At best, it might be possible to construct a new totally dynamical
>> routing protocol solely on a control theoretic basis as opposed to
>> the current protocols which are based in graph theory and static in
>> nature.
One of my coauthors wrote that particular section, so I don't want to
second-guess him. But I'd be awfully hesitant to say that ANY
approach is "the solution" until we have a consensus on the
requirements.
> >
> > Solution is ASM model:
> >
>> http://www.eecs.umich.edu/gasm
>
>Is this related to Dina Katabi's congestion control work in her
>thesis?
>
>http://ana.lcs.mit.edu/dina/dkpub.html
>
>Good stuff, but imo the main problem with applying control theory
>widely in the Internet is that traditional control theory isn't very
>good at being subverted and having the things it is using as control
>inputs _lie_ for their own ends. So, if you have state concerning
>utilisation in the routers (and state is expensive) tracking this -
>can you trust that state? It's always cheaper not to store the state
>and just make up a convenient lie on the spot.
>
>The problem with applying the abstract state machine model to the
>network is that you've really got a whole bunch of machines which
>don't (can't!) trust each other. (also why many bright-eyed ideas to
>load more information onto BGP data between ASs are doomed to
>failure...)
>
>The other problem is that you have to fall back to TCP congestion, as
>Dina's PCP does, to preserve the current anti-congestion collapse
>status quo.
>
>OTOH, right now we're simply trusting all endhosts to implement given
>heuristics effectively, which is equally flawed (Savage
>exploits, DOS attacks etc.)
>
>L.
>
>sure, it's possible. But is it useful?
>
><L.Wood@surrey.ac.uk>PGP<http://www.ee.surrey.ac.uk/Personal/L.Wood/>
This archive was generated by hypermail 2b29 : Mon Aug 04 2003 - 04:10:03 EDT