[gclist] Name that hypothesis

[Hans Boehm]

On Dec 3,  1:19pm, Charles Fiterman wrote:
> Subject: Re: [gclist] Name that hypothesis
>

> I would like to add a new hypothesis and emphasise the hypothesis part.
>
> 1. ...
> 2. The Collector may not need to know about the modifications. Henry
>    Baker has noted this in some of his papers.
>

Empirically, the collector has to know about VERY few of the modifications.
 The trick is to identify the right modifications.

Sometime early on in the development of our conservative collector I got the
sense of the dirty bits wrong, i.e. all clean pages looked dirty, and all dirty
pages looked clean.  Things would typically run correctly for seconds that way
...

Hans

-- 
Hans-Juergen Boehm
boehm@mti.sgi.com