From: Joubin <joubin@inch.com>
Date: Sat, 25 Oct 1997 09:35:21 +0800
To: "cypherpunks@toad.com>
Subject: Request for expert opinion and Feedback
Message-ID: <34514B35.AE826953@inch.com>
MIME-Version: 1.0
Content-Type: text/plain



(This message seems to have been dropped.  If you are recieving a
duplicate, I apologize.)



Hello all,

I am involved in a project which aims at the creation of a free and
portable Java(tm) based object operating system.  A basic functional
requirement for this new OS is to provide /fast and effective/ strong
cryptographic support at the OS level.  

The basic structure of the system is a layered architecture, with the
lowest layer consisting of a platform specific (non-portable) kernel
supporting an embedded JVM (again in C/ASM) and a set of Java(tm)
interfaces to this 'virtual platform'.  The OS proper will sit on top of
this and will be 100% Java (tm).

The issue:

As you may know, SMI has and will deliver a set of 'security' packages
as a standard component of JDK.  This uses RSA technology.  Basic and
general 'interfaces' to encapsulate constructs such as Key, Provider,
Cypher, etc. are also defined.  

A few people in the project group feel that many different groups and
entities out there will provide (portable) Public Key encryption
packages (implemented in Java(tm)) that can be used by the users of the
OS.  They argue against embedding computational support at the Kernel
(non-portable) layer as wasted effort.

Others (& myself included) feel that 'effective' also means efficiency
of execution.  Specially so if such functionality is to be used to
encypher 'streams' and such (if that is possible).


Your expert advice:

a) How do you feel about 'where' the computational support should be
   implemented ?

b) Assuming answer to (a) to be "At the kernel level and in C/ASM", is
   it possible to achieve a fine-grain modularity in terms of fairly
   generic algorithms and computations which can then be combined in
   conjunction to support a variety of encryption strategies ?

c) Is there any benefit to implementing the random number generation
   system in the Kernel?


Your input is greatly appreciated.  If you feel there are other
considerations to be made, please let me know.


Thank you.

Joubin

p.s.  Java(tm) and related items are the tightly guarded property of Sun
Microsystems Inc.

__________________________________
member, alpha zero LLC
joubin@inch.com
NoVA
__________________________________