On Tuesday, May 10, 2011 12:57:22 PM Sergey Dobrov wrote: > On 05/11/2011 02:44 AM, Justin Karneges wrote: > > On Tuesday, May 10, 2011 11:57:19 AM Sergey Dobrov wrote: > >> On 05/11/2011 01:46 AM, Justin Karneges wrote: > >>> On Tuesday, May 10, 2011 11:14:04 AM Sergey Dobrov wrote: > >>>> I think that it is enough to retrieve it by request with > >>>> node's metadata. > >>> > >>> The problem there is that you cannot track updates to the count without > >>> polling. > >> > >> Every event adds to the count one, am I wrong? :) > > > > Ah, so perform the query after an event happens? I guess that would > > work. > > Why? Query once, increment at each event.
This /almost/ works. Updated items that are not yet known to the client would incorrectly cause an increment. > Select count of items to often can be very inefficient, I think. Well, exposing the count would be optional. There are efficient ways to do it if the server cares. Justin
