In fact, at 10:30 AM, to be precise.
That's when I missed my first class of the year...
I flew in from La Guardia last night just to be here for it -- I
could have used a frequent flier ticket to return this morning after
class for free, but I paid $500 for a blackout date. To establish a
principle, turn a new leaf. To be here, in Irvine, present and
accounted for on the first day of classes.
But no, I slumbered until 10:46 AM in a masochistic proof that time
passes -- even if I don't :-)
I risked taking the car instead of walking over, though I don't have
a permit. I didn't know what classroom, to boot -- that info was on
the laptop, which was left discharged last night in a funk over abuse
from the cabdriver, who railed about being stuck with all these UCI
fares, wouldn't charge my ride (with only $5 in my pocket, I paid the
balance in dimes from my laundry coin jar :-), and then drove off
with my laptop still in the backseat. My roommate said it was quite a
ruckus, hollering and running after the cabbie down Palo Verde.
Anyway, returning to the classroom saga, I circled the CS building
where 90% of classes are held looking for Dr. Taylor, to no avail.
Now panicked it might be in some obscure corner of campus like
physical sciences, I went to Aileen to drop off my Orlando IETF
paperwork due a month ago (why? I've only been in California for
three days in December), but she didn't know either. So run up to the
fourth floor administrative offices to get the official course
catalog, run back down to CS 253 to... an empty room.
I don't mind that the first class must have been a short 15-minute
orientation as to the topic. That's natural enough. I don't mind
missing it -- there are always rationalizations that the info wasn't
*that* critical to enrolling or not (it's taught by my adviser, so I
will :-).
What bothers me terribly is that I've already blown my first
commitment to myself this year.
[Although, if I hadn't run back to my car just in time to see the
nice lady ticketing the car next to mine, the fine would have
*really* smacked me in the kisser :-]
----------------------------------------------
I'm planning for this to be my first year in graduate school -- while
I've been attending UCI since 1997, I don't think it's been my top
priority. So far, I've been leveraging the momentum I already had
going into it: broad involvement in the web standardization process
and a publication/speaking record all over the map.
I think I averaged a paper or talk once a week throughout 1998 --
burnishing my reputation, but at the expense of classes and
establishing a research program.
For 1999, my most visible resolution will be to hold official travel
to a minimum.
* I have one confirmed xmltalk, at WACC99 on Monday February 22nd.
* Hopefully WWW8 in Toronto in May -- I'll hear about the
evolution-of-w3c paper by Feb 1, and there's a panel on extending
HTTP Josh is organizing.
* IETF-Oslo in July is purely a wishlist item.
* I also proposed a Postel epigraph based on last year's IEEE columns
for INET'99 in June in San Jose.
That's it. I hope I'm learning to incorporate the lesson that I don't
have to accept every invitation or opportunity that crosses my desk
in order to be popular. I actually want to do some research.
First, though, I want to get my masters' degree by summer. There are
three major hurdles to cross:
1. Classes. I have let a lot of incomplete classes languish at very
low priority. They have now piled up too far -- even the registrar
isn't too happy. Papers on "ilities and protocol choices", "web site
design as configuration management process", "xml for hybrid
human/user interface description", and "online web help system
genres". I'm only taking one class this term, effectively, so that
leaves the rest of my week to tackle these. I want to be done this
quarter, by spring break. After that, there's one more formal methods
class requirement in the spring term.
2. Comprehensive exam. The only priority higher than classes is this
exam, based on about eight or nine inches of readings to become
familiar with. The exam itself is slated for springtime.
3. Publication. Not a full survey paper, which my
protocol-of-the-month columns are growing into, but just one
published, refereed paper that reflects the promise to do research. I
suspect something in the files or
coming up will suffice.
------------------------------------------------
Only then can I credibly embark on some heads-down research path. Of
course, research grant planning will proceed apace in the meantime,
but I have to admit to myself that it's a lower priority than closing
out the Masters'.
The general concept area is **software engineering for O(10^9) devices**
In particular I think I'll be developing a platform of sorts: an new
TP (Transfer Protocol) that routes information blobs within ad-hoc,
intermittently-connected, paranoid networks; and some insights into
related architectural styles. Following in Roy Fielding's footsteps,
so to speak.
Wireless is likely to be a salient characteristic of teracomputing
(not teraflops -- the domain of supercomputer wonks -- but trillions
of distributed digital devices (and also a fine pun on terracomputing
-- what Digital called Planetary Scale Systems like AltaVista)).
That's why I'm excited about UCI's proposals to consider how
event-notification protocols can be a useful style for wireless
applications.
The New York Times ran a year-end Outlook 1999 piece on the promise
of wireless to remap the entire industry. The point is not that the
article is newsworthy (NYT speculates on Apple's plans for their
$1500 consumer portable with CDPD this spring), but that the article
is appearing in the general press at all. I've included it at the end
of this message.
No, the much more galvanizing insight is that a little DARPA-funded
EE shop in Northern California is actually on track to build
Munchkins.
Aetherwire's "localizers" promise to be cheap silver-dollar scale
transceivers that can synchronize clocks with its neighbors to within
hundreds of picoseconds -- and thus positions within centimeters. The
point is not just to surpass GPS on the battlefield, but to pioneer a
new way of thinking about trillion-node networks: geographic routing,
paranoid security, and nomadic connectivity. I've attached my
analysis from two months ago.
So there's the question I want to set to: how can we program terracomputers?
Client-server alone isn't the answer. At some point, there isn't any
one email or web or news server for one user's professional life. The
application, as well as the data is just "out there". This isn't the
same as parallel computing -- little computers in formation may form
big ones, but the normal mode of operation won't be forks and joins,
I think.
Happy New Year,
Rohit Khare
-------------------------------------------------
Date: Mon, 23 Nov 1998 12:36:21 PST
To: gorlick@aero.org, fielding@ics.uci.edu, ejw@ics.uci.edu,
dsr@ics.uci.edu, taylor@fdr.ics.uci.edu
This weekend, I spent some time with the PIs of a DARPA program for
"Warfighter Visualization" out of the Electronics Technology Office
who have been developing a 'localizer.' It's a disposable
ultrawideband transciever that can approach group clock accuracy
within tens of picoseconds. That is to say, they can triangulate
positions within centimeters or better. The effective bandwidth is
1-5 kilobits/sec. It's currently the size of a business card,
intended to scale down to a quarter.
[Another political tidbit: Tom Kalil was there from the White House
and we discussed the latest BAAs; he said that "addressless"
networked systems were a buzzword he'd been told to expect from
Tennenhouse; it sits neatly astride 'deeply networked systems' and
our own insight that 'IP host addresses' are not the right way to
distribute events.]
http://www.darpa.mil/eto/
"The Electronics Technology Office focuses on electronics technology
to produce smaller, lighter, and more mobile information systems."
http://www.darpa.mil/eto/wv/index.html
"Warfighter Visualization provides unique ways of receiving and
interacting with spatially correct, timely information in both
individually served and shared environments."
http://www.darpa.mil/eto/wv/98overviews/aetherwire.html
"To develop totally integrated ultra-wideband (UWB) transceivers for
communication and position location. These "Localizers" will be
sufficiently small, low power, and low cost to be used for multiple
military and commerical applications. They will allow position
location to centimeter resolution over ranges of hundreds of meters.
Localizers will operate within a network of hundreds to thousands of
units in a local area, including inside buildings, urban areas, and
forests.
http://umunhum.stanford.edu/~ss/aetherwire/pi_rep95.html
"Aether Wire's long term goal is the development of coin-sized
devices that are capable of localization to centimeter accuracy over
kilometer distances. These "Localizers" will be able to operate
within a network of millions of other units in a local area, and
users will be able to enter and leave the network seamlessly and
transparently. Ultimately, these localizers will be able to operate
for up to a year on a watch-sized battery, or longer if augmented by
solar power. The overall goal of this ARPA sponsored project is the
development of pager-sized units powered by AAA-sized cells that are
capable of localization to submeter accuracy over kilometer distances
in networks of up to a few hundred Localizers."
[I'll refrain from commenting on the EE bits, which are fascinating.
Ultrawideband is sufficiently different from sine waves and even
spread-spectrum that the only way I can grasp it is as "sculptured
noise"]
[I saw a more recent PI slide set from Sept 17 98]
Lead .mil users: IFF, Logistics/warehousing, ground surveillance,
mine clearing, smart minefields.
Lead .com users: personal location, inventory, IVHS (cars), robotics,
home networks
ANALYSIS: these folks are waaay at the bottom of the stack and they
know it. Their ambitions are clearly in the right place, but
they/their grant doesn't have the software to deploy applications
significantly more interesting than position and sensor telemetry.
ETO/AFRL has been actively disinterested in their software components
(beyond the software radio and their positioning simulator).
This kind of platform raises lots of fun dynamically reconfigurable
architecture issues. [...]
The pager platform is a very reasonable way to ramp up to this
environment: it emphasizes the wide-area, low-bandwidth distribution
of tactical information to nodes with significant local compute power
(AW's localizers have a Motorola ColdFire and 2Mb onboard; they're
only using about 70Kb).
What this adds is a backchannel, ranging even up to satellite
integration (a cloud of such devices could form a phased-array
antenna...); and accurate enough physical ranging to consider
"geo-routing" rather than packets.
The heart of the WISEN proposal (that event semantics can be
exploited beyond packet semantics alone) is still true. I'm excited
because this platform motivates a second component: that message
semantics can be exploited beyond packet semantics alone, too. That
is to say, these devices *are* munchkins: they use 'currencies' to
arbitrate local bandwidth, have enough cycles and bits to store and
forward and manage entire mails/posts/pages/event queues, and form
ad-hoc 'paranoid' networks.
Seeing this project has broken a logjam in my thinking. I really
believe these devices are going to happen, even without
nanotechnology. In fact, they're going to happen in the timeframe I'm
expecting my PhD. I want to understand how to write decentralized
applications for these things ("physically realizing C2
compositions").
I'm going to try and finally assemble the argument I've been making
in bits and pieces over the last three years: that IP is not the way
to build emergent applications. Markets are part of the answer;
application awareness/adaption to distribution is another; and open
protocols which subsume Internet information services are a third.
In fact, I simultaneously resoved something else to myself (just in
time for New Year's): no more travel. My only 1999 engagement is
WACC, and I'm going to limit myself to papers I want to publish, not
invitations or requests.
Besides, I need the time for Phase IIs and old classes...
Rohit
--------------------------------------------------------------
January 4, 1999
Still Waiting for a Computing Vision to Materialize
By JOHN MARKOFF
PALO ALTO, Calif. -- In Silicon Valley there is perhaps no greater
mystery than why it has taken anytime-anywhere wireless computing so
long to arrive.
Wireless data, after all, is not a new concept among the computer
cognoscenti who frequent the coffee bars, restaurants and bookstores
that line the avenue leading to the Stanford University campus.
Indeed, the idea was sketched out clearly as long ago as 1972 by two
Xerox researchers, Alan Kay and Adele Goldberg, in a never-realized
concept for a portable computing device they called the Dynabook.
>From the start, the Dynabook was intended to allow its owner to
compute and retrieve information while sitting under a tree in a
meadow.
For many years, the more enterprising of Silicon Valley's hardware
hackers have built their own experimental wireless systems.
There are even some promising signs in the consumer marketplace. For
example, two-way pagers with tiny keyboards have recently gained some
popularity, and the widely popular Nokia 6160 PCS phone being offered
by AT&T Corp. has the ability to receive short e-mail messages.
Wireless computing is, in short, a universally shared vision. Yet, in
the immortal and slightly skeptical words of the former Byte magazine
columnist Jerry Pournelle, its realization is still scheduled to
arrive "real soon now."
The problem, depending on whom you ask, is either that a number of
barriers remain, ranging from low bandwidth to the lack of compelling
applications, or that the Valley's visionaries have simply been out
of step with consumers.
"We're still in the snooping-around phase of wireless data," said
Jerry Purdy, president of Mobile Insights, a Mountain View, Calif.,
computer and communications consulting firm, who is one of those with
the former view. "We're still trying to get it right."
Others are more pessimistic.
"This is a zero-billion-dollar industry," said Geoff Goodfellow,
founder of Radiomail, the first wireless electronic mail company.
Goodfellow left Radiomail after failing to create a broad consumer
market, and the company is now owned by Motorola Inc. Renamed Blue
Kite, it is attempting to offer specialized wireless services for
corporate applications.
For evidence that wireless data may be a harder sell than the
industry's optimists believe, consider the travails of Metricom, the
wireless Internet company that has deployed an innovative network in
the metropolitan areas of Seattle, San Francisco and the District of
Columbia.
Now owned by Microsoft's co-founder, Paul Allen, Metricom offers
unlimited access to the Internet for $30 a month via its Ricochet
radio modem, which is the size of a deck of cards and offers a speed
of 20 kilobits a second. Though that is much slower than today's
fastest modems, which transmit data at more than 50 kilobits a
second, the Ricochet's dependability and wireless convenience have
earned it wild praise from many of Metricom's subscribers.
Yet this high satisfaction and three years of aggressive marketing
have netted Metricom only about 20,000 subscribers nationwide, and
Allen will soon be forced to inject more cash if the company is to
survive.
Even so, despite the disappointing markets for Radiomail, Metricom
and even AT&T's cellular digital packet data network, which allows
data transmission over telephones, there is no shortage of fresh
initiatives to bring on the wireless digital future.
Earlier this month, Palm Computing, the division of 3Com Corp. that
makes hand-held computing devices, introduced, with great fanfare,
the Palm VII, a shirt-pocket-sized product that will offer instant
access to Internet data and e-mail.
Scheduled to be available sometime this year, the device will sell
for about $800, and data access will cost about 30 cents for 1,000
characters.
Its backers argue that the Palm VII will succeed for the same reasons
the original Palm won a devoted following -- instant access to
information, in this case via the Internet, without the hassle of
having to boot up a full-scale personal computer.
"The ability to get information without waiting is really important,"
said Donna Dubinsky, a Palm co-founder who left the company early
last year to found Handspring Inc.
Palm's announcement is likely to be followed soon by a similar
wireless Internet data announcement from Steve Jobs, founder and
interim chief executive of Apple Computer. The computer industry is
keeping a careful eye on Apple's plans because the company has a
demonstrated knack for coming up with user-friendly software and
services.
Several industry insiders said they believed that Jobs had struck a
deal with AT&T to use its wireless data networks to connect portable
and desktop computers to the Internet. Neither company will comment.
Still, even Jobs, who is widely admired in Silicon Valley, may have
his work cut out for him.
For one thing, the proverbial "killer application" of wireless
appears not to be e-mail. Anytime-anywhere messaging is essential for
a few on-the-go high-technology workers but apparently not for the
average consumer, who has steadfastly ignored any wireless text
messaging system larger or more expensive than a pager.
Still, the optimists insist that the realization of wireless is only
a question of time and that all the drawbacks will vanish with the
advent of third-generation broad-band wireless data networks
scheduled to go online in 2002.
In an industry long accustomed to successfully marketing technologies
that people didn't know they needed, the optimists are gambling that
the ability to send megabits of data to a hand-held gadget each
second will eventually generate the elusive killer app.