From owner-freebsd-hackers  Fri Mar 31 04:15:10 1995
Return-Path: hackers-owner
Received: (from majordom@localhost) by freefall.cdrom.com (8.6.10/8.6.6) id EAA17710 for hackers-outgoing; Fri, 31 Mar 1995 04:15:10 -0800
Received: from deacon.cogsci.ed.ac.uk (deacon.cogsci.ed.ac.uk [129.215.144.7]) by freefall.cdrom.com (8.6.10/8.6.6) with ESMTP id EAA17688; Fri, 31 Mar 1995 04:14:48 -0800
Received: (from richard@localhost) by deacon.cogsci.ed.ac.uk (8.6.10/8.6.9) id NAA10050; Fri, 31 Mar 1995 13:09:43 +0100
Date: Fri, 31 Mar 1995 13:09:43 +0100
Message-Id: <199503311209.NAA10050@deacon.cogsci.ed.ac.uk>
From: Richard Tobin <richard@cogsci.ed.ac.uk>
Subject: Re: HotJava 
To: Bakul Shah <bakul@netcom.com>
In-Reply-To: Bakul Shah's message of Thu, 30 Mar 95 11:51:00 -0800
Organization: just say no
Cc: Dave Cornejo <dave@dogwood.com>,
        jkh@freefall.cdrom.com (Jordan K. Hubbard),
        freebsd-hackers@FreeBSD.org
Sender: hackers-owner@FreeBSD.org
Precedence: bulk

> You need OS support to turn a data page into an executable
> page (usually an executable page is read only & shared).
> mprotect() should do the trick though.

Data pages are executable anyway in FreeBSD.  Numerous program already
load in compiled code without needing to call mprotect().  But in
general, yes, some OSes have this feature.

>  Also, AFAIK, not
> every processor allows you flush cache via user mode
> instructions so you may need OS help there.

True, but so far we only have x86 to worry about :-( and it doesn't
have separate I/D cache.

> Scheme would be ideal if it had a
> decent interface to arbitrary C functions + an on-the-fly
> code generator.

Hmm, I wonder if the Java bytecode language is adequate to support
Scheme (it has multi-threading, so even call/cc might be possible).
Since it's the bytecodes that are downloaded, you would be able
to write Scheme programs for HotJava.  I'll ask on the Java mailing
list.

-- Richard