Digital Media

Moshell -Spring 99

Lecture 25: Multimedia over Communication Networks
and
Information Appliances

This lecture concerns two topics. Wu Chapter 9 describes how MPEG-2 can be transmitted via ATM to constitute a Video on Demand service. (We discuss this topic in order to inspire you with visions of entrepreneural glory, wiring up hotels on I-drive or something.) Then, to keep the audience awake, we will engage in a design exercise concerning "information appliances" or, as Wendy Kellogg of IBM puts it, "Virtual Reality inside out."
 

Video on Demand (VoD)

By now we've already discussed the failure of the Full Service Network experiments here in Orlando. However, VoD is a booming business in hotels, where the operative word is "Captive Audience". Folks on travel are often under stress and desperately need entertainment, and they need it NOW, not whenever HBO gets around to showing whatever they want to show. Hotels are in the business of providing what you need NOW, expensively.

VoD in an ideal version would allow the user to select any movie and start it any time. The basic architecture is a hierarchy of caches with the Core Service Center (CSC) having the original copies of the movies, in a compressed format of course.

I have read the first half of Wu Chapter 9 (pages 301 to 314) without finding a single interesting fact. Everything they say makes sense, but it's mostly names of conventions and services.

Time Stamping is somewhat interesting, on page 315. The problem to be solved is to deliver video at a constant rate. Even though ATM can guarantee QoS, it is easier to specify a slightly higher QoS than needed, than it would be to maintain perfect synchrony between transmitter and receiver (like analog TV.)

When the text mentions a "reference timestamp at 27 MHz", this means that its numerical accuracy is capable of specifying 27 million different times in a given second (or to put it another way, an accuracy of 1/27 of a microsecond or about 37 nanoseconds) - not that they transmit 27 million timestamps per second. Likewise the deconding timestamp has a resolution of 90 kHz.

Query 25.1: In an NTSC video signal with 525 lines * 30 frames per second, to what spatial resolution does 27 Mhz correspond? That is, what fraction of a horizontal line of video represents the variability resulting from a 37 nanosecond (1 bit) timing error?

Query 25.2: On page 316 the expression phase lock loop is used. What is a PLL and how is it used to establish a local time reference, e. g. for the chroma signal in an analog TV system?

Query 25.3: On page 315, there's a diagram that shows a constant delay from end to end of the MPEG-2 system timing model, with variable delays internally. Why do these variable delays occur? Is there anything surprising about WHERE the delays occur?

Query 25.4: Who cares about synchronous broadcasting anyway? Why don't we just record some amount of the signal and then generate our own local clock, reconstitute a composite video signal with chroma and all, and not even worry about how fast, exactly, the transmitter is sending the information? (two issues: uplink cost, and "The Sting", a famous Redford movie.)

Error Correcting Codes are an extension of the checksum concept. It's a slow lecture, so let's look at ECC for a bit.

Checksums, you will remember, provide information that allows you to detect an error. For instance a parity bit can detect any single error in a word (including an error in the storage of the parity bit itself.) But you don't know where the error occurred, and so you have to request a retransmission. A simple two dimensional scheme that provides error correction is to have each byte contain a parity bit, and to also have a longitudinal parity checkword after each block of perhaps 16 bytes. An error can then be detected by finding the intersection of the bad row and bad column.

Here's a link to error correcting codes.

Information Appliances

Wendy Kellogg, a visionary researcher with IBM,  told a story at the 1991 Cyberspace Conference. She said that a successful technology comes to you, you don't have to go to it. Putting on a Head Mounted Display and gloves, entering a wierd low fidelity simulated space, was the paradigm of "going to it." Her vision of cyberspace "coming to her" was a story about her kitchen. It went something like this.

When she wakes up in the morning, her coffee pot, knowing her habits, has already made the coffee. As part of its routine checking around, it noticed the night before that there was not much cream in the fridge and so it mentioned to the shopping-list in Wendy's PDA, to add that to her shopping list. It also checked to see if the toaster was prepared to make toast, but the toaster was not functioning. So the coffeepot (being the toaster's buddy, after all) phoned a repair shop and placed an order for a service pickup. The coffeepot checked with the household butler to schedule the pickup at a time it knew Wendy would be home.

And so it goes. Nothing very high tech is happening, but it's all INTEGRATED.

How could these appliances talk to each other? What kind of standardization is needed?

Information appliances are those whose primary business is handling information: PDAs, pagers, wristwatches, celphones, computers. The Palm VII is a nifty example. Coffeepots and other gadgets that should be better informed than they are, could be called smart appliancesor perhaps  informed appliances.

Sun's original vision behind Java was as a programming environment to solve problems like these. The second generation of thought on this subject has led to Jini, a protocol for information appliances to talk to one another. Please read this article and answer the following queries.

Query 25.x1. List the first five devices you interact with every day. List information that each of them would benefit from having (i. e. would serve you better if they had the information.)

Query 25.x2. Describe several ways in which an informed appliance could find out whom it was dealing with (human, that is, who doesn't have a Jini interface.)

Query 25.x3. Describe several ways in which smart appliances could be used against their owner. What kind of safeguards are possible in principle to protect against such threats?

Back to the course index
Back to the course syllabus
Back to the previous lecture
Onward to the next lecture