Internetworkers: http://www.ibiblio.org/internetworkers/

[Ron Thigpen <rthigpen AT nc.rr.com>]

zman wrote:


> It's still a big winner in my book.

right on.  esp. considering the timespan for getting this done, the fact 
that they were using a new technical approach, and the low cost at which 
they delivered.

to mind it's unfortunate that this is being viewed by many platform fans 
as a proxy battle for their preferred hardware.  makes about as much 
sense as "my chevy is better than your ford" based on NASCAR outcomes. 
whatever.  it's good tech either way.

and that add'l testing and tuning is a delay of sorts, but it's yet to 
be seen what performance improvements that will yield.  they were 
getting ~80% of theoretical efficiency in an earlier test w/ fewer 
nodes.  the whole cluster only yielded ~44%.  this may be a true blue 
problem in scaling this architecture, or it could be due to a bottleneck 
that can be eliminated in the coming weeks.  let's hope for the latter. 
 it makes for good watching in any case.

--rt