Microsoft's Tiger Unleashed [ I do not know the author so cannot evaluate the quality but .. djf]

[David Farber <farber () central cis upenn edu>]

Date: Thu, 19 May 1994 15:04:25 -0400
From: discover () halcyon com (Mike W. Perry)




In Seattle on Tuesday (5/17/94) Microsoft unveiled its new media server
software developed under the code name "Tiger." I attended the unveiling
and thought Roundtable participants might appreciate a report, particularly
from the perspective of what it means for "grassroots" organizations and
independent producers.


The technical description for Tiger is "continuous-media software server."
Like ordinary file servers, it supplies users with the contents of files
stored on an array of hard disks (generally, video digitized as MPEG-1 or
2). But Tiger is designed with the particular needs of video-on-demand.
Given a request, the typical file server dumps the file from its own hard
disk to the user's PC as fast as possible. This approach would not work
with video. Either the transfer would go too fast, overwhelming any storage
buffers the viewer might be using or it would go too slow, causing the
picture to pause and jerk. So Tiger has to be smart enough to release the
frames of video at about the same rate the viewer displays them, hence the
"continous-media." It also has to be able to do typical VCR functions like
fast-forward and pause.


Microsoft claims Tiger will be an open standard running on a wide variety
of platforms from high-end PCs to something Intel calls their "scalable
multi-servers." With current data buses, Microsoft claims a single Pentium
PC with the proper video cards could service about 30 viewers, providing
each with any one of dozens to thousands of videos (depending on hard disk
capacity) and enabling each to pause, fast forward etc.With the new Intel
high speed bus they claim that same PC could handle up to 100 viewers,
enough for many businesses, hotels, hospitals and the like. Cable systems
will use either high-end file servers by companies like Compaq (1000+
viewers) or Intel's multi-servers (tens of thousands of viewers).


In the media, Microsoft and its partners, Compaq and Intel, are clashing
with companies like Oracle and Silicon Graphics who claim that video file
servers need much more powerful and expensive computers as well as highly
sophisticated software. At the moment, I can't say who is right,
Microsoft's demonstration featured only 16 monitors though it performed
well enough with them, not even missing a lick when one hard drive and
server PC was turned off. (The system is supposed to be highly fault
tolerant.) Microsoft's approach, if it works, will have an economy of
scale. Media servers will come down the same assembly lines as millions of
PCs and use chips already in mass production. Since I think "cheap" is one
of the keys to a successful information superhighway, I applaud their
effort and hope it succeeds.


Among the gurus, there's a debate whether the average home will access the
information superhighway ("Super-I") though something that's more like a
TV, with TV-type remote controls and screen menus or something more like a
computer (alpha keyboards, a mouse and more approaches to user input).
Microsoft sees the future being like a TV and the viewer interface to
Tiger, what little they showed, reminded me of an "on screen" VCR. I'm
skeptical how well this will work in situations where the user will be able
to choose from thousands of films and shows--the result would be an endless
series of confusing menus or long, page-down lists with all the advantage
going to those who get listed near the top. In this kind of situation, a
computer-like interface that would let you do a "find" on movies titles,
actors, keywords or whatever would actually be easier to use. Once
interesting movies were found and placed in an individualized on-screen
list, a TV remote control would then be fine after a viewer settled back in
his recliner. At any rate, the point is probably moot. If Microsoft makes
the system as open as they promise, third parties should be able to market
all kinds of interfaces and set-top boxes to talk to the Tiger server. The
system won't be proprietry, even Tiger will be clonable.


One of the developers told me that the videos that the viewer watches do
not have to be stored locally on the server, that Tiger can bring in video
(with some delay) stored on distant machines with no additional effort on
the viewer's part. This should help independent producers compete with the
handful of blockbuster films that will undoubtedly be stored locally.If you
can get your work somewhere on the cable system, anyone on that system can
view it. (Now all you have to do is get on the menu.) However, this ability
is not like ftp on the Internet, the viewer can't simply go anywhere in the
world and find anything. The cable company still controls what the viewer
can access and isn't likely to let viewers roam to far. The only real
change is that the scarce resource has changed from air time on a limited
number of channels to space on a limited number of hard disks and hard disk
capacity is easier to expand.


There are opportunities here. Because the system works with all forms of
media that can be stored on hard disks and not just the usual TV
broadcasts, non-profit organizations could produce a kind of video or audio
magazine that the cable system could receive cheaply overnight (when the
Super-I is lightly loaded) and viewers could select and watch just as
easily as they do CBS Evening News. The key would probably lie in getting
enough viewers to demand space for the video magazine on disk. Similiarly,
a "movie of the week" club could be run the same way. If enough viewers
request and are willing to pay, the cable company would be under pressure
to put the video on disk for viewers to watch at their leisure.


A developer told me that the control signals that viewers can send to Tiger
are not limited to VCR-like commands, that they can be expanded to include
commands for complex games or virtual reality. There's potential there for
the creatively inclinded. There's also the opportunity for online
instruction and even a virtual school that responds to student input.


Microsoft has teamed with TCI to test Tiger in Seattle and Denver in 1995.
They claimed they intend to move quickly from a technical test using
Microsoft/TCI employees to reaching all TCI's Seattle and Denver service
areas. Full scale marketing will begin in 1996 with the world release
following quickly on the heels of a US release. With much of the hardware
already in mass production, volume production could begin quickly. That is,
of course if Tiger works as they claim.


Feel free to post or published this elsewhere.


Mike Perry, Discovery Institute
<discover () halcyon com>