[thelist] Interupting the holy wars for a moment

Steve Lewis slewis at macrovista.net
Fri Jun 14 15:30:01 CDT 2002


Dunno, maybe this belongs on thechat but here goes.

Mark Groen wrote:

> I got the same thing as Jon. Packet sniffers such as the ones they are talking
> about are kinda worrisome to me also as it's expensive enough as it is for
> broadband for those in the under 30K crowd. Although, a lot more people

My first reaction was, "oh, someone has been sniffing around in IPv6
documents again."  (see the 8 bit Traffic Class field in IPv6 headers)

Then I remembered that ATM networks do something like this; ATM has 4
service models, with different QoS implications, and different cost
structures: CBR, VBR, ABR, and UBR (which stand for constant/ variable/
available/ unspecified bit rate).

In the case of ATM and the priority flags in IPv4, the pricing
differentiation has not been used because the processing cost of these
accounting services were too high, and degraded router performance too
greatly.  That may be changing however.

The problem is that the networking equipment must go all the way to the
application layer, the deepest nested comms protocols we use, to
differentiate this traffic.  This equipment will be expensive because:
teaching a device about all the comms protocols in the internet
increases cost.  Inspecting the innermost protocol means traversing the
protocol stack to reach the bottom and then performing the analysis,
which means performance would suffer as each packet is "held-up" for
analysis before re-transmission, and so the equipment will require more
computational power and the analysis logic will require extensive tuning
by developers (something the article seems to overlook in it's
statements about 4GL languages decreasing time to market) so overal
device performance can reach an acceptable level.

Finally, while the stated goal is "to build networks that will be able
to properly authorize, authenticate, and account for customized services
by identifying individual packets and classifying them based on
individual customer usage and QoS requirements," this seems to assume
that content encryption will not occur at the transport layer (TCP/UDP)
or above.  If it does, than content classification will fail because
these are end-to-end protocols, where encryption is encoded at packet
creation, and (idealy) can only be decrypted by the recipient.

Ppresumably they would charge their highest rate for any encrypted
content, including HTTPS traffic.  That will encounter more than a
little difficulty taking hold in the industry, and would probably
require <US centric>"illegal collusion" on the part of the comms
companies <rant>(not that the current administration would
notice)</rant></US centric>.

In summary:  The possability is more than a little frightening, but
Orwell would be proud.

--Steve




More information about the thelist mailing list