RE: Platform Independence (was George is Farr from where we want to be)

I believe platform independence is very important to IBM and probably
also very important to ISVs but of little to moderate importance to the
end user community.  For IBM & for ISVs it leads to products that can be
deployed across the various platforms they sell (IBM) or choose to
support (ISVs).  It lets them market this ability to not be locked to a
specific platform.

One downside for hardware & OS vendors is that platforms are
commoditized.  For IBM this must be frustrating.  They need to market
cross-platform technologies to compete but at the same time those
technologies gut the mindshare that makes companies go for their
proprietary (System i & z), high-margin platforms over lower cost,
higher bang-for-buck (System p & x) hardware.

Another downside is that cross-platform code is rarely optimized for any
specific platform.  This frequently leads platform choice to cost,
speeds-n-feeds, internal preferences, and strength of the OS and
available tools/utilities.  Companies are getting used to horizontal
scaling out (more small systems) v. vertical scaling up (few bigger
systems).  Again this favors Windows & Linux over midrange.


Beyond the hardware and app vendors, the user community will generally
pick a platform and stay with it for a number of years, making the
independence issue less a factor.  It is a factor when selecting a
software system and will influence the platform decision, but in the end
companies rarely rip and replace one platform with another just because
the apps support it.  There has to be a business reason.

A sample business reason is when an infrastructure upgrade is necessary
to support a growing or changing workload.  There are more upgrade
options and more technologies to consider.  I've mentioned before that
we're a JDPeopOracle EnterpriseOne shop and that our i5 is not sized to
handle the growth we're going through.  As IBM has killed upgrades
within the original 570 frame (1.65GHz CPUs) the upgrade option is to
gut the CEC and replace it with the 2.2GHz CPUs.  I'd like that, of
course, but that means re-buying our RAM and misc. other expenses like
$50+K/CPU core for OS licensing.  Upgrading both production & BCDR is a
high 6-figure expense without even adding any DASD.  The alternative we
are piloting is moving part of the architecture -- the end-user piece
served by WebSphere App Server -- off the i5 and onto Windows.  

So the cross-platform nature of WAS and EnterpriseOne's use of that
technology is losing revenue for IBM as the Windows boxes we buy come
from Dell.