tag:blogger.com,1999:blog-17924780.post1293424677737400082..comments2007-10-29T10:52:19.656-07:00Harlan Iversonhttp://www.blogger.com/profile/14403505969348821916noreply@blogger.comBlogger2125tag:blogger.com,1999:blog-17924780.post-88475645570251374582007-10-29T10:52:00.000-07:002007-10-29T10:52:00.000-07:00Cool, interesting point. So, the direct reference doesn't have to be zero, but there's a sort of indirect count that could be zero if a hierarchy were not connected to a known root, similar to modus ponens.<BR/><BR/>A (and only A) connects to B, B and C connect only to each other and/or A; if A becomes eligible for GC (no strong references from a known root), then by implication B and C are also eligible for GC.<BR/><BR/>Is that about right?Harlan Iversonhttp://www.blogger.com/profile/14403505969348821916noreply@blogger.comtag:blogger.com,1999:blog-17924780.post-64787254712833147632007-10-18T11:52:00.000-07:002007-10-18T11:52:00.000-07:00Hi Harlan,<BR/><BR/> I just wanted to comment about circular references, in Java, and any language that uses some form of mark-and-sweep GC algorithm, circular references are not an issue.<BR/><BR/> The Java GC starts from known 'roots', most notably the execution stack, and crawls through all references marking objects as live. All objects not marked get swept away. So, even your circular list will be destroyed if it's no longer accessible.tcrimihttp://www.blogger.com/profile/01872736152611370009noreply@blogger.com