Graham Leggett wrote: > William A. Rowe, Jr. wrote: > >> It seems that a more effective high-water mark is to set an upper bounds >> on the desired cache size. > > My understanding is that working out the size of a directory is more > expensive than working out what you have left on a filesystem. Is this > correct?
Your understanding would vary widely by the filesystem in place. HFS, NTFS, ufs, etc etc all have different performance penalties for this computation. If you are ANYWHEREs near filling the same volume as temp, you are already likely screwed.
