The library looks pretty interesting, and I've been looking at the
interface and white papers to see how it was written.
Good white paper:
http://cache-www.intel.com/cd/00/00/30/11/301132_301132.pdf
The part I don't get?
Its C++.
They missed the point. Most of open source is in C. If you want
ubiquity, you write in C. You can wrap C into C++.
Or Perl, Python... but C++? Pain in the ass.
If I write something for just myself I'll use C++, but if I want
ubiquity? Its going to be C.
What would have been an even bigger leap in thought for this release?
Find a way to incorporate the library in Apache's APR or in the GNOME
libraries.
The second part that they missed?
Its GPL, not LGPL, not BSD, but GPL. They have a waiver on it, but it
would have been better to just say it was BSD.
Licenses are business models. I have a hard time believing that Intel
needs to make money off this library. The GPL creates an environment
for dual licensing. Great business model, but from listening to the
Intel speakers, this is not what they were aiming for,
They seem to be pushing for ubiquity. What is the license model for
ubiquity?
BSD. The BSD license provides for the opportunity for ubiquity.
The interfaces look good for Intel's Building Block Library, which
means there exists now examples for these algorithms that can be
followed, so this is a gain for open source.
I just wish the library would have come in a package that was a bit
more useable.