Jon Skeet said:
<snip>
And this is an important point. IIRC, Longhorn will have three
rendering modes - one for slow machines, one for fastish machines which
can't quite afford to do *all* the swish stuff, and one which has the
works.
If Longhorn is released in 3 years' time (just a complete guess) then
according to Moore's Law, CPUs will have 4 times as many transistors -
and graphics cards at least *have* been obeying Moore's Law squared, so
they'll have 16 times as many transistors. That's a *lot* more power to
chuck stuff around the screen.
Computers which are now pretty cutting edge will be at *best* the low
end of what you can buy, and there'll be far fewer 500MHz machines out
there to contend with. (I doubt whether Longhorn will even claim to run
in 64Mb though.)
Basically, Microsoft can afford to make display stuff reasonably
expensive, because there'll be better hardware to chuck at it - the
balance of what's best for the developer vs what's best for the
hardware will have tilted significantly in favour of making it easier
for the developer.
Yes, I agree, however thats part of the argument against porting it back to
say, XP. I expect XP to run at 500mhz at 64meg(I have a machine running at
just that actually). This kind of rendering engine would be required to be
capable of working on all machines the target system supports, otherwise its
pointless, and properly ratcheting up or down depending on hardware adds
complexity to the design and increases the costs of development and testing,
which is what I was trying to get at. 3 modes * 3 systems(2k, xp, longhorn)
* minimum configuration sets is bound to grow quite a lot, increasing the
amount of testing(and probably code) required to get it working across
systems in a reliable manner. I don't think any system component should be
made that changes the system requirements, that is something best left up to
the next version, IMHO.
Anyway, the point is, what use is Avalon on a system that is too slow to use
Longhorn? If the user is willing to upgrade his machine, why not his OS,
which with new hardware is generally not too expensive? Even in corporate
environments, I doubt administrators would take a shine to Avalon any faster
than Longhorn itself, considering the amount of OS level modification
required to use it. Also, what use is Avalon on systems that don't have
WinFS or any of the other features. Avalon alone is not enough to make me
want to move to Longhorn, it is simply another piece of the puzzle. I don't
think I'll be developing for Longhorn using JUST Avalon, but I will be
attempting to leverage everything the platform offers. Suggesting that all
features should be backported seems absurd to me...it seems that people
would rather turn XP into Longhorn, which is, in my mind, exactly the
purpose of upgrading.