How Fast Can a Simple Implementation of E Be?

[Mark S. Miller]

Announcing the ENative Project.

At http://www.erights.org/enative/ I've posted web pages and source code 
for a very simple and readable implementation of Kernel-E in C++.  As 
explained on these pages, were this work taken even a moderate amount 
further, it would tell us, rather reliably, how fast a very simple 
implementation of E in C++ along these lines would be.  Ping provoked me 
into doing this (thanks!), by showing that the current interpretive E on 
Java is much slower than Python, while reminding me that Python has a very 
simple implementation that people are nevertheless happy with.  I see no 
reason that an E implemented at the same ambition level cannot do at least 
as well, and I suspect it could indeed do somewhat better.  ENative is my 
incomplete attempt at finding out.

The implementation techniques used come from Lisps, Smalltalks, and especially 
Joule.

With the posting, I hereby put the work down for now.  Hopefully one of you 
will pick up where I've left off.  If you're at all like me, I promise you 
it would be a fun project.  I found it very hard to stop working on it.


         Cheers,
         --MarkM