[mythtv-users] Mooting architecture for a DataDirect replacement

[Jay R. Ashworth]

On Wed, Jun 20, 2007 at 11:58:09AM -0700, Chris Petersen wrote:
> We are also looking at other providers, as well as the option of working
> with other projects (both open and closed source) to band together on a
> similar service.

That seems like a launching point for this discussion.  Since
architecture is the only thing I'm really good at, I'll start.  :-)

(If you're not interested, Kill the thread.  If you're offended, relax.)

If the time comes when we have to create a service to replace Z2L's
DataDirect, there *have* to be better ways to accomplish it, leveragint
technologies and protocols which have already been invented in the last
30 years on the internet.

To decide which ones might be best, I think we need to start by
enumerating the goals and describing the data and it's sources.

The fundamental item -- and yes, I know, DD's schema is floating around
in (I think) a PDF -- is an AIRING, which is a combination of PROGRAM,
NETWORK, and TIME.

NETWORK maps to CHANNEL through SERVICE, TIME includes a TIMEZONE and
is exact to the second, and PROGRAM is, where possible, primary keyed
by PRODUCTIONCO and PRODUCTIONNUM, which I believe to be globally
unique.

The PROGRAM records originate from the production company -- for
Scrubs, frex, that's Touchstone TV.  The AIRING records originate from
the network, and filter through the local station, for network
affiliated broadcast stations.  For cable, they mostly don't filter.

One of the *most* important characteristics of AIRING data for the PVR
market is that you want to propagate changes to the audience just as
*late* as you possibly can -- 5 minutes before airtime is not
unreasonable, nor impossible, if you pick the right transmission
architecture.

This will occasionally screw with MythTV's schedule architecture, but
that's the price you pay for predictive scheduling.

The second most important issue is *accurate* start and end timing.  If
you're gonna start ER at 9:59, damnit, *say so*.  There's no reason to
get that wrong except passive aggressive attempts to piss off PVR
owners enough to watch live, IMO.

*My* personal approach to the transfer problem would be to package up
whatever updates a given source has, and post them to a Usenet
Newsgroup on some server, appropriate signed with a public key so it's
clear that they're valid, and possibly encrypted with another key, if
that's necessary commercially.  A system similar to CSS could be used,
where a random symmetric key is PKC encrypted with multiple public
keys and appended to the packet.

Each box could then subscribe to that newsgroup, and ping the server
for new news every ... 5 minutes?  The infrastructure is *designed* for
that sort of thing, and very well understood; proper server coverage --
even in these days of Supernews et al -- would be trivial, and flood
quite quickly... and you could push out "The president's on!  He's on
*every channel*!!" updates in jig time.

Thoughts?

Cheers,
-- jra
-- 
Jay R. Ashworth                   Baylink                      jra at baylink.com
Designer                     The Things I Think                       RFC 2100
Ashworth & Associates     http://baylink.pitas.com                     '87 e24
St Petersburg FL USA      http://photo.imageinc.us             +1 727 647 1274