Twobit or not Twobit
Mark S. Miller
markm@caplet.com
Wed, 30 Aug 2000 14:44:38 -0700
At 02:22 PM 8/30/00 , Ben Laurie wrote:
>... Amen! Would it be heresy to suggest that C++ may be a more
>appropriate target?
If so, can I be both heretic and high priest of the E religion? ;)
http://www.erights.org/enative/index.html is such a proposal. For many
reasons, I am going to continue investing my efforts primarily in E-on-Java.
But I posted these pages because I'd love to see E-on-C++. The posted
pages show a mapping that is small, beautiful, and reasonably efficient.
Together with the existing Kernel-E documentation, this is also enough
information for someone to pick up the ball and run with it. I now await
volunteers.
Following a suggestion by Tyler, if we define Core-E as
1) Kernel-E,
2) the user-syntax to Kernel-E expansion, and
3) a core set of libraries (yet to be separately defined) adequate for
running headless confined smart contracts,
then, for reasonable engineering effort, Core-E-on-Java and Core-E-on-C++
can be compatible. The largest piece of engineering would be porting
Pluribus (the org.erights.e.net.* packages). These should be ported to E
rather than C++ so that one source can be maintained for both platforms.
Since E as a whole hijacks the Java libraries, a compatible complete
E-on-Java is only feasible when combined with a Java native interface, so
that the generated C++ code may still make use of these extensive Java
libraries. Note that I don't care if Python or Squeak libraries are be
better. They aren't the same.
Cheers,
--MarkM