|
|
Excavating the past, for a better future. | |
|
WHITEBOARDS links into revise activity how to use SITENAV meetings what's new? whiteboards post article frontpage downloads ORIENTATION legalisms history glossary participants BACK-ENDS udanax-green udanax-gold ALGORITHMS coordspaces enfilade ent OLD MANUALS XIA HELPING puzzles needs funding site-traffic admin
|
Copied from a forum message Objects migrate to and from the disk in groupings called "flocks". The flock is the minimum granularity at which objects are purged from core and restored from disk. A flock consists of exactly one shepherd and an arbitrary number of sheep (as long as the entire flock is guaranteed of fitting in a single snarf? Is this true? Michael? Dean? How can we be confident this condition is satisfied?). A "sheep" (what is the singular?) is simply a "COPY(X):" object which needn't be at all cognizant of the disk. The single shepherd in a flock is "representative" of the flock as a whole--the only inter-flock pointers are to shepherds. This means that the only on-disk pointers to a sheep must be from objects within the same flock as this sheep. If a sheep is an EQObject there is no problem, because a hash table that points to it must be within the same flockD. Given that hash-indexed structures rehash when being read back in off disk, the flock will always be using valid hashes during its stay in core. --JohnDougan? |
|
| All trademarks and copyrights on this page are owned by their respective companies. Comments are owned by the Poster. The Rest ©2001 Jeff Rush. |