Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.1 6/24/83; site ariel.UUCP
Path: utzoo!linus!decvax!microsoft!uw-beaver!cornell!vax135!ariel!jlw
From: jlw@ariel.UUCP (J.WOOD)
Newsgroups: net.unix-wizards
Subject: Re: Berkeley Flame
Message-ID: <518@ariel.UUCP>
Date: Sun, 13-Nov-83 07:53:37 EST
Article-I.D.: ariel.518
Posted: Sun Nov 13 07:53:37 1983
Date-Received: Mon, 14-Nov-83 01:37:50 EST
References: <591@ihuxx.UUCP> <18957@wivax.UUCP>
Organization: AT&T-ISL, Holmdel, NJ
Lines: 26

Steve Dyer's article makes the same semantic mistake that I
have seen in many articles recently.  The specific statement
made is that BTL UNIX doesn't have 'virtual' memory for its
VAX version.  Of course it has virtual memory; what it doesn't
have and what all the submitters mean is that BTL UNIX doesn't
implement a paging virtual mamory system.  BTL UNIX implements
virtual memory meaning that the address bit pattern generated
in an instruction or by some arithmetic operation in a register
for example is translated by a memory management unit before
a main store reference is made.  On the other hand, when BTL
UNIX needs some more main store, it selects a process to be
deleted and ships the whole image to a swap device.  When
Berkeley UNIX needs some more main store it looks for a page
to delete.  This is more efficient than the BTL way.  The
other alternative which is becoming more attractive for
many sites is to just buy enough memory.  That runs faster
than either.



					Joseph L. Wood, III
					AT&T Information Systems
					Laboratories, Holmdel
					(201) 834-3759
					ariel!jlw