I am working through strategy/architecture for a distributed project and some of the nodes will not be connected to the Internet from time to time.
So … one gets into the whole area of managing the replication, object id’s, timestamps and so on. I realize that products like iAnywhere/ASA and Sybase MobiLink do a superb job of this sort of thing.
HOWEVER, now I am beginning to wonder about management of the Primary Keys in an SAAS environment in zero-bandwidth cases.
Does anyone have any thoughts on Servoy managed sequences / auto-enter versus a db managed sequence if deploying with remote nodes (some stations connected, others will be traveling in zero-bandwidth mode)? Experiences here?
Tentatively, I have been reflecting on a standard primary key (say INT or BIGINT) along with a Remote_ID or Node_ID. But, no point in re-inventing the wheel either. Regardless, I am wondering how this will all fit with a Servoy managed sequence.
When you go with iAnywhere or anything else where you would create records outside of Servoy you should always use database sequences.
I think, when the answers start floating in, the majority will be in favor of db sequences.
If it comes to replication, you should NOT use numeric IDs. If you do that you will have to create different number areas for the sync clients. In the world of replication, the best is to use UUIDs, since they are unique in any case. I have already filed a feature request that asks for a new sequence option that fills UUIDs. Most DB vendors are capable of UUIDs, but I suppose the downside is that they are all implemented differently. So I’d prefer Servoy to manage that.
I think you are mistaken. As long as I understood replication, this is the only stratigh forward approach. Everything else involves thoughts on who gets which IDs etc…