Here is testimony from David Reed at a hearing today. Please read it – not only is it both excellent and thorough, it’s also only ten pages long.
David says (paraphrase mine – read the testimony) that because of the inappropriate surprise, disruption, and risks created by DPI, it shouldn’t be allowed. (Or we should move to all-encrypted communications.) He also makes the key and constant point that the internet has flourished because transport was, traditionally, predictable. This allowed innovative new protocols to be born. (This battle can sometimes be framed as a right-to-life tussle on behalf of as-yet-unborn technologies.)
Of course this tradition of predictable transport didn’t start with the internet. The challenge for the entire argument against DPI is to tie the meaning of predictable transport, and its contribution to economic growth, to a workable legal regime that politicians will be willing to endorse. Arguments from history are powerful (and there are certainly many of them here). Arguments for the future are harder, and David Reed’s contribution here is powerful.
Implementation is tricky – in this case, a “no DPI” approach gets involved (potentially) with national security, child porn, and copyright concerns. At the least, “no DPI by network operators for their own commercial benefit” has to be the rule. And for those other problems, legally-authorized (with probable cause) access to individual data streams has to be a better idea than inspecting every last packet.
This story is only just beginning.
“At the least, “no DPI by network operators for their own commercial benefit” has to be the rule. ”
This appears to prohibit network operators from selling anti-sparm, anti-virus, and many other things people currently buy (or bundle) as part of their Internet service because network operators have a commercial benefit.
David Reed’s testimony is motivated by his “end-to-endian” dogma, which states that there should be no intelligence anywhere in the innards of the Internet. (Never mind that Internet backbone routers are, in fact, special-purpose supercomputers!)
The truth is that there is really no such thing as “deep” packet inspection. A packet of data is one-dimensional; it has no depth. (Think of a postcard.) Every part is as visible as every other, and any “depth” is only in the mind of the reader.
The word “inspection” is also deceptive, because it implies that a human is looking at the packets. In fact, a machine — which cannot “comprehend” the data — is doing relatively simple pattern matching, looking for patterns that might suggest spam, or malware, or illicit activity such as P2P. Such is no more invasive of privacy than, say, an automatic “sniffing” machine that detects potentially hazardous materials in physical mail.
In short, the whole “DPI” brouhaha is really just another tempest in a teapot created by would-be regulators of the Internet.
Susan,
Your concerns seeem to largely echo the excellent policy concerns David raised, but there was another point made at the hearing as well. As was pointed out by the Chairman calling the DPI hearing, and echoed in the comprehensive testimony provided by CDT’s Alissa Cooper – never mind the policy debates, there’s state and Federal law prohibiting the practice.
I share your concerns as a mater of policy, but why doesn’t the wiretapping/Cable act/state law illegality of this practice (assuming inadequate/no consent) get equal attention?
Good point, Ethan. I’m writing a post this Saturday going into the DPI question in greater detail, and I’ll try to answer you there.
I realize that I’m commenting on old news, sorry. Nonetheless: I find Dr Reed’s analysis interesting, but not universally applicable. There’s two things that are hard to buy: One is that simply pushing more packets faster is the best solution in every case (in my experience, it isn’t, due to hard technical limits that exist in some setups), the second one that DPI equals modifying data in-flight. While some systems do that, it’s an overly broad generalizations to say that all do.
That is not to say that the man doesn’t have points. He does. But they’re not (necessarily) applicable to all the gear he says it is applicable to.
If you can live with a short bit of advertising, I wrote a longer entry about this on my own blog.
http://www.shortpacket.org/2008/08/on-dr-reeds-dpi-hearing.html