Path: utzoo!utgpu!watserv1!watmath!att!rutgers!apple!bionet!arisia!roo!mark
From: mark@parc.xerox.com (Mark Weiser)
Newsgroups: comp.object
Subject: Re: Precise GC (was Re: Do we really need types in OOPL's?)
Message-ID: <572@roo.UUCP>
Date: 18 Oct 90 05:16:11 GMT
References: <736@tetrauk.UUCP> <18261.27131385@qut.edu.au> <563@roo.UUCP> <1990Oct16.015230.24169@cbnewsc.att.com>
Sender: news@parc.xerox.com
Lines: 31

In article <1990Oct16.015230.24169@cbnewsc.att.com> lgm@cbnewsc.att.com (lawrence.g.mayka) writes:

>I was referring to "exact" garbage collection, which collects *all*
>inaccessible storage.  My application must run continuously over long
>periods of time without leaking memory.  *Any* degree of conservatism
>is too conservative for me (in the matter of GC, anyway).

I am not sure you read Hans carefully.  One of his points was that
there is no such thing as exact.

With regard to running continuously over long periods of time: of
course.  We have many servers written in Cedar with this
characteristic.  Cedar has used partially conservative (from stacks)
collection since 1982 or so, and fully conservative since 1988 or so.
All is well.

And it is important to distinguish kinds of leaks.  Conservative
collectors tend to leak once and then stop--because some magic number
in the data or stack has held an object inappropriately.  But this is
not a growing leak.  Always just that one object.  There are some
cases to be careful of--if that one object is the head of a long growing
linked list you were hoping to be collected you could be in trouble.
Our practice here is to explicitly zero the references out of linked
lists as we are done with them.  

-mark



--
Spoken: Mark Weiser 	ARPA:	weiser@xerox.com	Phone: +1-415-494-4406