Manuel Jesus Petit de Gabriel <freston@xxxxxxxxxxx> wrote on Sun, 6 Jan 2002 18:22:43 -0800: > Allowing attributes to be written incrementally has some performance > issues (each write would force the node to be reindexed.), some > boundary conditions (writing at offset=8 when the attribute only had > 2 bytes... how should it be indexed), etc. It never was high on the > requirement list to devote extra time; there was no compelling reason > for implementing it given the intended usage for attributes. The API > however was defined for allowing it, if at a later time it was really > required. Yes, that makes sense. Only allow one bulk write so that the index doesn't need to be updated. I was thinking about using a file handle to the attribute (well, actually an attribute handle), and updating the index when no open handles remained. But that's not the way BFS went, and it doesn't seem to have hurt it much. - Alex