On Thu, Aug 2, 2012 at 9:59 AM, Alexander Shorin <[email protected]> wrote: > On Wed, Aug 1, 2012 at 12:44 PM, Robert Newson <[email protected]> wrote: >> The attachment data would be streamed, not fully materialized into memory, >> so the difference is the size of an attachment chunk (which is under 8k >> iirc). > > That's right for client and direct view request, but I'm asking about > _list processing on query server side. Streaming attachments to it via > stdin wouldn't solve that problem if attachments will be embed into > doc object. If not: I suppose then there would be new command set that > breaks compatibility and it's not clear how to fast read and process a > lot of chunks within fixed os_process time. > Ah. you're right lists processing would be a problem with the current implementation. But imo that is another problem that could be fixed later. Current implementation of the query servers can be improved a lot...
- benoît
