Re: 64-bit transition deadline (Re: Etch in the hands of the Stable Release Managers)
On Mon, Apr 09, 2007 at 02:06:14PM -0500, Manoj Srivastava wrote:
> > I think we agree that Moore's law predicts the available amount of
> > memory.
>
> Broadly. Precision, however, is not something Moore ever said
> of his statistical law,
The "perfect" prediction you want does not exist. But Moore's law has proven
very reliable in the past, and I don't see why it won't be in the following
2 years.
Just look at the graph, e.g. for intel cpus:
http://www.intel.com/technology/mooreslaw/pix/mooreslaw_graph2.gif
> so you can't base decisions to the precision of
> months based
Decisions taken by humans (and specificaly, decisions taken by the release
team) are always based to some extent in imperfect predictions. I don't see
why this should be different.
> solely on that.
Nobody was saying that this should be the _only_ matter in consideration. Do
you have any particular reason not to release in ~18 months ?
> > High-end application developers will always write software that
> > accomodates to that amount, so when this amount is 8 GiB,
>
> Then again, these decisions are made based on future market
> share and prediction of future user bases on part of the application
> developers.
No, they're based solely on the fact that in an highly competing market, the
application vendor who only takes profit of the first 4 GiB will lose to the
one who takes profit of the full memory available.
> Secondly, these large application developers you seem to speak
> of appear to be proprietary application developers -- which mean that
> in my eyes the issue pales to insignificance. Are we talking about
> closed source software here?
We're talking about software in general, be it free, propietary, or even
software written for private use. Neither the application developers have to
be "large" nor the software has to be non-free. It's a small niche, so why
not small, GPLed vendors?
I would love if it could be filled with free software only, because that would
imply that all non-free win32 applications would go away with microsoft. This
is, however, outside of the scope of this discussion.
> > There won't be a "big" migration at that time, but the decision of
> > which will be the reference 64-bit platform will be taken and set in
> > stone.
>
> I personally find it very hard to credit that an initial
> decision based on insufficient data and projected user bases is going
> to be set in stone.
The key is network effects. When DOS was set in stone, not even Microsoft
was able to replace it with Windows untill they started offering the huge
advantage of native 32-bit addressing and protected mode.
--
Robert Millan
My spam trap is honeypot@aybabtu.com. Note: this address is only intended
for spam harvesters. Writing to it will get you added to my black list.
Reply to: