The FCC has officially endorsed the digital television standard
proposed by the "Grand Alliance," so now is a good time to explore
exactly what lies in store. DTV represents the complete convergence of
the television and computer industries, but the standard itself is
almost entirely a product of computer design paradigms - much of the
specification is in a remarkably C-like pseudocode (including most of
the semicolons). DTV is incessantly described as an open, flexible
standard; it is this claim that I will dispute most directly.

I have discovered among the DTV documents promulgated by the Advanced
Television Systems Committee (ATSC) the following remarkable facts,
which surely represent some of the most profound contributions to
sociology, perceptual psychology, and information science, that this
century has seen (note sarcasm).

-There are exactly 256 languages in the television-viewing world.
-There are exactly 65536 program sources in the world.
-Cinema-quality image trains can be compressed with a ratio of greater
  than 38:1 with perceptually unobjectionable results.
-Only two aspect ratios exist: 4/3 and 16/9
-Subband-based video compression is universally appropriate.
-Audio compression by frequency-domain quantization is universally
  appropriate.
-There are exactly four types of recording studios in the world:
  none, large, small, and reserved.
-Five channels of audio are sufficient for any program - six are
  supported by the bitstream standard, but the header allows a maximum
  configuration consisting of L, C, R, Ls, and Rs.
-Eight types of audio program exist: "complete main," "music and
  effects," "visually impaired," "hearing impaired," "dialogue,"
  "commentary," "emergency," "voiceover," and "karaoke." Test and
  calibration signals are perhaps "music and effects?" Second-language
  dubs are perhaps for the "hearing-impaired?"

DTV specifies MPEG-2 as the sole algorithm for image train
compression.  It is possible to supply a packet stream to the raw
transmission subsystem using an alternative algorithm, but only a
correspondingly customized receiver would be able to decode the
signal, and it is not at all clear whether the FCC would permit such a
practice in broadcast-based systems without a lengthy and costly
approval process.

Incontrovertibly, the specification and enshrinement of a particular
pair of algorithms for video and audio compression is symptomatic of
classic computer engineering short-sightedness; as with the millenium
bug and the 32 bit internet address, a moment's careful consideration
would reveal the folly. A thoroughly satisfactory solution would be to
allow the program supplier to include the decoding portions of a pair
(audio and video) of companding codecs of its own choice (and
perhaps design) as software to be run by the receiving device. Work
on such systems is already well underway; hardware capable of
decompressing 1920x1080x30fps video (DTV's highest quality) in better
than realtime using a virtual machine approach will be cost-effective
by 1998, when DTV is slated for its first non-trial deployment.

Despite its overprecise specification of content attributes (including
ratings, specifically MPAA and Expanded EDS) and compansion
strategies, DTV makes only the briefest mention of "scrambling" and no
mention at all of signatures.  Considering the quandary in which
Washington politics have mired cryptography, this is hardly a
surprise, but the oversight is not thereby made in any way excuseable.
If parents are to be empowered such that they are confident they can
prevent unauthorized viewing of destructive programming in their
households, programming must not be considered suitable unless the
programming, with attached ratings, is digitally signed. Once all
programming is in the digital domain, with the means of distribution
increasingly complex, such capers as Captain Midnight's (who hijacked
HBO in the '80's) will become more common, not less. Consider how many
web sites of stolidly reputable (or at least stolid..) organizations
have been cracked and doused in ample doses of pornography.

One final glitch of note: I found no mention of the Viterbi FEC
algorithm in any of the documents. The RS-interleave-Viterbi-
interleave FEC pipeline has become a de facto standard for through-
the-air digital data delivery, and even with this powerful FEC
strategy DBS systems still suffer from frequent catastrophic EC
failures. High-throughput Viterbi codecs are ubiquitous and cheap, and
are used in many if not all modern hard drives, so I don't understand
the omission.  I am not optimistic that DTV, with data rates
substantially higher than those of current DBS systems, will be
reliably delivered through the air with only RS, interleave, and
clever trellis coding.

The official documents specifying DTV and its subsystems can be found
at ftp://ftp.atsc.org/pub/Standards. See particularly a_52.pdf,
a_53.pdf, and a_56.pdf (.rtf and (wheeze) MS Word .doc versions also
there).