[IP] Re: Scnneier on Crossing Borders with Laptops

[David Farber <dave@farber.net>]

________________________________________
From: Rod Van Meter [rdv@sfc.wide.ad.jp]
Sent: Sunday, June 15, 2008 6:05 PM
To: David Farber; marcaniballi@gmail.com
Subject: Re: [IP] Re:   Scnneier on Crossing Borders with Laptops

On Sun, 2008-06-15 at 10:30 -0700, David Farber wrote:
> ________________________________________
> From: Marc Aniballi | Personal [marcaniballi@gmail.com]

>  there is nothing wrong with FedExing your Memory Stick or even
> your whole notebook bag to your destination. Then you could take
> a good book on the flight and relax, knowing that whether or not
> it arrives, you won’t have to lie to any government officials!

Not really practical.  Even sending it in a package, it has to clear
customs -- and in practice, is probably *more* likely to be tied
directly to *you* than if you hand-carry it, because a package
*definitely* generates a paper trail that most luggage carried through
the airport doesn't.  People who deal in exotic, one-of-a-kind
prototypes are often happier hand-carrying them than sending them, and
worrying about the physical safety of the device is only one reason.

Also, even today, FedEx can take as much as a week to send a *letter*
from the U.S. to Japan, because it has to clear customs when it gets
here, and customs is most decidedly not in a rush.  Do you really want
to send your laptop out a week before your Tokyo meeting?  (DHL is
slightly better in Asia, FWIW, but customs is the real variable.)

>
> I’ll send it ahead. Or I’ll bring empty hardware and send the
> content ahead – or download it when I get there.

I have a student who's interested in this.  But a quick calculation
suggests it's an iffy proposition:

A planeful (500) people of people, each carrying, say, 100GB of data
(call it 1Tb = 1E12 bits for round numbers), taking ten hours
Narita-->LAX, that's 5E14 bits/36,000secs = ~10Gbps.  Oh, and the
NRT-->LAX "pipeline" holds a couple of dozen flights, so you're into the
hundreds of gigabits/second.  Even intelligent deltas and prioritization
of certain types of data still leave it as a herculean task.

(When I worked for the MOSIS project, back in the Dark Ages of
nine-track tapes, we used to worry about which had a higher effective
bandwidth -- a FedEx truck of tapes, or FTP.  Tape won, hands down, but
the latency is high :-).)

As long as I'm talking, for those who (like me, to a large extent)
studied congestion control in grad school, but haven't followed it
closely since, you might like to check out
http://www.icir.org/floyd/longpaths.html

There are a number of projects there, including Steven Low's Fast TCP,
which is now a startup in Pasadena, selling an appliance that
accelerates TCP without getting bad marks for "plays well with others",
by measuring congestion in a very different way.  (Presumably the
appliance behaves as a TCP proxy, though I haven't checked.)

                --Rod





-------------------------------------------