[gclist] zillion objects

[Ji-Yong D. Chung]

Hi

    Thank you for your suggestions. 

    If I may summarize: part of your solution is to divide up the 
oldest generation space into separate partitions and 
perform collection on one or two of them at a time, using 
remembered sets.  Andrew Chen also has suggested
similar approach (though he did not spell this out as
explicitly as you have).

    I like the simplicity of the solution.

    Perhaps I can try the LRU or MRU as criteria for 
doing the collection over the partitioned old data.