Pretty straightforward question here: when setting up one-way replication, what are the pros and cons of choosing Child-to-Parent versus Parent-to-Child in an all-SDE environment? The documentation doesn't provide much guidance. I can't discern if one has any advantages (or disadvantages) over the other as the differences don't seem to be relevant to SDE. (I gather that older ArcGIS versions required edits to from Parent-to-Child (is this correct?), though that no longer seems to be the case.)
A single SQL Server "editing" SDE (10.1)
A single SQL Server "publication" SDE (also 10.1)
The SDEs will not be used for disconnected editing, hosting geometric networks (or similar), and there will only be a handful of editors. This is a very simple setup where we wish to replicate an editing SDE to a publication SDE. The publication SDE is used for publishing services, serving out layers, etc. We don't intend to replicate with file or personal geodatabases.
That’s what I originally thought—but then I saw this in the 10.3 documentation:
“One-way, child-to-parent replication works in a similar manner, but in the opposite direction. Here the data in the child replica is editable, but the data in the parent is considered read-only.”
So given our configuration, I don’t know why I’d choose one over the other.
I guess it doesn't matter then; I started using replication at v 9.3, and recently migrated from 10.0 to 10.3.1, and have just kept with what works for me. In my simple mind, the parent (edits) to child (read only) seems to make logical sense.