[gclist] What to say about GC.

[Jerry Leichter]

[Story of GC bug that lead to an obscure problem in a symbolic algebra program - 
"perhaps this is what gave GC a bad name".]

For every story of such a GC bug, one can probably find 50 stories of bugs with 
similar effects in compilers, especially in optimizing compilers.  (Have a look 
at comp.compilers for a recent thread of examples.)  I doubt there's been a GC 
written that's a tenth the size of a *small* modern optimizing compiler.  Yet, 
this hasn't "given compilers a bad name", and I don't recall the last time I 
heard someone argue for using assembler to avoid all those compiler-caused 
problems!

I used to use SNOBOL4, one of the earliest GC'ed languages, quite extensively.  
I've never heard of a bug in the GC algorithms of any of the three different 
implementations I've used over the years.  Ralph Griswold's book on the original 
"MAINBOL" implementation discusses problems that arose, and as I recall makes no 
mention of GC bugs.  Oh, I'm sure they existed in early versions, but then so 
did other bugs, which could have just as serious - and just as subtle - effects.

(As an interesting side note, SNOBOL was originally developed as a language in 
which to do symbolic algebra....)

I'm preaching to the converted on this list, of course, but somehow the message 
has got to get out....
							-- Jerry