Testing the waters.

Kelly Murray kem@Franz.COM
Thu, 08 May 1997 22:32:07 -0700

> engineering to separate out platform-dependent code from
> platform-independent code. If LispOS is based on CL, application level
> code should be VERY portable between LispOS and other CLs. Only the

This is exactly correct, I see probably 90% of the code for this
LispMachine to be portable CL code.
Another 8% will be the locks, transactions, versioning,
that is specific to the persistent memory.

The remain 2% will be hardware/underlyingOS specific.

This means for 98% of the contributions to this LispOS effort,
using Persistent-webserver-ACL-on-Linux should not be an issue
as compared to Unwritten-persistent-wrong-approach-CL-HTTP-not-fully-capable-yet-CMU-CL-on-low-level-Fluxtoolkit,
and why I've been saying we can make lots of progress quickly
if we don't try to reinvent the Persistent, the HTTP, the ACL,
and the Linux part over again before making foward progress.

The CMU-CL-on-Fluxtoolkit can come later, much later.
In fact, I believe it is best to delay that work until 
most or all of the whole thing is already written and working.

Since, at that point the effort to clone persistent-webserver-ACL-on-Linux
would add some value if for whatever reason ACL wasn't acceptable,
because Franz charged 99cents for it, or it wasn't GPL'd,
or all the source wasn't available or whatever.
Or the Linux kernel WAS GPL'd or it was written in C or whatever.

98% of the working system would already be done,
and the rest would be well understood and well documented and proven
as useful.

-Kelly Edward Murray