Gilder, worth repeating:
from the April, 2001 letter:
"Corvis concretes, Ciena softens
If the hardware innovations at OFC offered widespread confirmation of the
PowerMux powerdigm, the intellectual case made a leap as well. Unexpectedly, it
was
the enigmatic Corvis who advanced the new paradigm in optics most eloquently and
clearly. Speaking on a panel with Calient's Tim Dixon, Ciena's Steve Chaddick,
Chorum's Scott Grout, and representatives from MEMS makers OMM and IMM,
Corvis' chief hardware engineer Dave Smith issued a challenge to the
conventional wisdom of optics.
The discussion focused on the software needed to control all the lambdas and
datapaths. With perhaps the leading all-optical cross-connect, Calient shifted field
ominously and now claims its real value is in network management and software.
Uh, oh. At the forefront of developing a wavelength analog to multiprotocol label
switching (MPLS), Calient's Dixon reminded us that 50 percent of his company's
engineers are in software, as all future hires will be.
Chaddick stressed the difficulty of the software problems Ciena's Core Director has
overcome and insisted Calient's purely photonic cross-connect wouldn't meet
the needs of carriers. "Forty percent of our customers' interfaces are still STS-1s (52
Mbps), another 40 percent are STS-3s (155 Mbps), and only the remaining
20 percent are split between OC-192s (10 Gbps), OC-48s (2.5 Gbps), and OC-12s
(622 Mbps)…If you don't have a hybrid switch with electronic interfaces, you
don't get to sweep the Nortel gear out of the network." Challenging Dixon's claims
of software expertise, Chaddick added, "Anyone that doesn't have a switch
through trials with a carrier doesn't know how big the software problem is….There
is this monstrous control plane issue to deal with."
Confidently, if subtly, Dave Smith suggested a very different approach. "There are
many ways to build an all-optical network, but it's not a given that you have to
have a tremendously dynamic core. What we need are architectures that give
control at the edge," he asserted. With the continued doubling or tripling of data
traffic
every year, electronic switches will not suffice.
"Who says we are going to have some arbitrary limit on the number of wavelengths
in a network? The number of wavelengths we want is not determined by the
capacity requirements. Even in a tiered network where we have, for example, just
16 nodes, you need enough channels to connect all the nodes….Architecture
drives the number of channels to fully connect your network." Smith wants to waste
bandwidth to multiply scarce connections. Methodically, Smith finished his
dissection of the Soft Network advocated so often by Ciena, Sycamore (SCMR),
Calient, and even Cisco (CSCO). "I would focus a word of caution on software.
Some of the biggest crashes in telecom history have been the results of software
failures….look at your PC and see how often if crashes….I do not trust handing a
multi-terabit network over to software engineers."
Paul Green said years ago that with fiber optics the quality of service (QOS) issues
so dear to electronic switch and software makers would go away. Light would
be 10 orders of magnitude more reliable and 10 orders of magnitude faster. The
network would harden. Whether in the BlueArc Silicon Server or in a Corvis or
Avanex network, software complexities rooted in the scarce processing power and
memory of the microchip can be driven out by wasting the abundant gates of
FPGAs or the copious bandwidth of optical fiber.
Today the network remains largely hard on the outside, with TVs and telephones
and appliances, and soft on the inside with code rich routers and switches and
protocol converters and add-drop mazes.
Soft on the outside, hard on the inside, the new network will put the intelligence on
the edge where the people are and the hardware where the photons are. It will
put electronics where memory and processing is needed and glass where the speed
of light is the only constraint. Thus, the network will conform to the physics of
light and electrons and to the needs of its users on the Net. This is the promise of an
inspiring OFC, where the all-optical network emerged at last from the vapors
and assumed the imperious reality of thousands of arduously designed and
manufactured systems. While the market still is saying no to the new network
topology,
OFC issued a resounding yes.
George Gilder and Charles Burger
April 5, 2001"
and then there is this:
Instead of behaving like day traders, successful inside investors buy when
their companies' stocks have attractive valuations, due to external conditions.
Then they hold on, waiting for a rebound they're confident (based on their own
assessment of the company's outlook) will come. "If they've been around long
enough, they know how the cycles go and which time of year the stock tends to be
cheap," Gerber explains.
Technology
It was with trepidation, however, that we looked at the tech stocks generated
by our screen. As we noted, several are profitless and many are in troubled
markets. Gerber points out that tech insider buying is especially noteworthy,
because the sector is awash with stock options. When executives have ample stock
options (granted at discount prices), buying shares on the open market shows
twice the commitment, in Gerber's view.
So what about the buying at money-losing Corvis (CORV)? The February purchase
of 150,000 shares for $12.85 a pop by David Oros, who sits on Corvis's board
(and is also the chief executive and chairman of Aether Systems (AETH)), seems
admirably risky. But so far, it looks a little questionable as optical-equipment
sales slump. The stock has been slashed in half to $6.05 in just six weeks.
Plenty of insider selling has followed suit. But because neither Oros nor his
Corvis peers have been buying and selling for long enough, there isn't a track
record to turn to."
Draw your own conclusions here. I did and bought this stock |
