Re: Another IDS Feature Request

Art S. Kagel said >> 1) I'd like to have a pre-existing statements and declared cursors so that >> my application can just attach to them. >That was what someone else asked for (peraps boson?), I want/need: That's "Bozon" to you buddy. ;-) as in: > Bozon - It is the elemental particle of stupidity of which all other > stupidity is composed. This particle is indivisible and cannot be > broken down into any smaller parts of stupidity. http://groups.google.com/group/comp.databases.informix/browse_frm/thread/0de5ebbc0ff2a8b2?hl=en Yes, it was I who requested that. The thread then started riffing on two different ideas: Replicated connections and replicated cursors. Replicated connections of course would be fantastic but I would be happy for replicated cursors. So we started with a simple idea of replicated cursors and then when we got some momentum we just completely overran ourselves with the notion of virtual connections. My earlier post today on implementing replicated cursors also reduces the communication. When the driver is ready for data it tells both (or more) cursors that it is ready for data and takes data from the first cursor that provides it. It then sends a signal that it has the data it needs. Each cursor continues to get ready for the next request. I don't care if replicated cursors are limited to read-only. The biggest problem with this for ER is heterogeneous replication where each server has vastly different data. You could solve this a couple of ways. 1) disallow it. Don't allow global cursors if any of the tables in the query are not completely replicated. 2) embrace it. Basically the query becomes select * from <replicate1> union select * from <replicate2>. This would allow developers to do some "interesting" things. Of course don't give any of the developers working on this my name or address. Non of this sounds easy.