[mythtv] Whats wrong with myth

[Brad Templeton]

On Sat, Mar 19, 2005 at 09:35:51PM -0500, J. Donavan Stanley wrote:
> Don't take all of the above to mean it's all been sweetness and sunshine
> here in Stanley manner.  I've had my share of headaches, but my
> headaches have come from experimentation and me doing boneheaded things
> to one or more of my machines for the most part.  The worst experience
> I've had was that whole "unreadable file handle on file creation"
> problem that appeared sometime around .15 or so. (Incidently the new
> Suse MBE doesn't have this problem)

Of course all problems we encounter have a cause, and often it is our own
error or ignorance of a procedure or failure to read certain docs or running
something unusual.   But those causes don't mean a zero-incident install.
Linux has come a long way of late.  There's a lot of software now that you can
run just by "apt-get install" or a GUI to it.  And that's good, very good.
I don't expect Myth to be at that level yet, but I was surprised at how many things
could go wrong.  This is because there are so many pieces, and many of them are
young.  People like Tivo don't have this worry, they control the platform and
once the first developer gets it working it works for everybody.  Making complex
software work on all the variety of platforms is hard work.  (Not that it isn't
hard for Microsoft either.  They control the OS -- sort of -- but actually end
up having to support even more cards and software packages than linux devs do.)

That it's a hard challenge however does not mean it's not a worthwhile one.
With MythTV, we could write up a description of a standardized box, dictating
the video card, capture cards, motherboard chipset and linux distro and saying,
"Here, if you buy exactly this hardware, and install everything in this big
package, it will work."   We could even do better than that.   But it's not the
linux way, is it?  We want to run on every distro, every system but the most
strange.   You get people saying, "I want to run myth with <bizarre card X>" and
we try to answer it, rather than saying, "Spend $35 and get the card people have
worked hard to debug on."  Amazes me sometimes.

Why does all this matter?  PVRs are one of the "important" technological products
of the last decade.  They are cool, and now they have enough buzz that everybody
wants them, though many don't yet know why.  Myth's the most advanced linux pvr,
and so people are highly interested in it, but to bring most people in you do
need an experience perhaps not as plug and play as a Tivo, but perhaps on a par
with installing BeyondTV or MCE -- or gimp.   We can't deliver that if you want
to run any hardware at all, though it could happen with a reference implementation.

There's a lot of stuff the open source community will probably never do because
it's a lot of unrewarding legwork. I can imagine a system that you install and
it looks at your feeds for closed captions and other data and checks other systems
to figure out what cable system you are on, and configures itself for that.  A
system that hardly even asks any questions but works.  That would get pvrs --
and linux -- in millions of new places.