[gclist] Re: gclist-digest V3 #2

[Richard Jones]

>Date: Sat, 30 Jan 1999 12:45:24 -0500 (EST)
>From: John R Levine <johnl@iecc.com>
>Subject: [gclist] Computing Surveys on distributed GC
>
>The current (Sept 98) Computing Surveys has a long article on distributed 
>GC, that is, GC of objects where the references may be in different 
>address spaces separated by slow network links.
>
>Anyone else read it?  Whaddaya think of it?

I've only read the draft version that's available from their web site; the 
Computing Surveys version may have improved on this.

The paper is an extended version of their IWMM'92 paper. I don't know what the 
lead time for Computing Surveys is, but this might account for Eliot's 
disappointment that more recent work is not included. However, for my money, 
the Plainfosse and Shapiro DGC survey in IWMM'95 was a substantial improvement 
on either of the Abdullahi et al surveys.

I didn't find this survey very convincing. Too much of it was taken up by an 
account of centralised GC - a subject much better handled by Paul Wilson's 
surveys - and I missed any substantial account of the factors that influence 
the design of distributed garbage collectors. The draft also contains errors 
(for instance the summary tables 1 and 2 at the end confuse "direct" and 
"indirect") and at least one of the algorithms cited (Dehne-Lins 1994) is 
broken but the draft survey does not mention this.

Of course, these matters may heve been rectified in the published version.

Richard Jones