Keep data objects' identifiers on the identifier table on CNs

[taojing2002]

trafficstars

Currently on the CNs, the data objects' identifier are not stored on the identifier table since CNs doesn't synchronize the data objects. Chris thinks we should keep them there since the identifier table is the only place we hold all object's identifiers. The systemmetadata table is not the place since the delete method can remove items from there.

[mbjones]

I agree.

[taojing2002]

We are going to remove the identifier table on 3.0.0

[mbjones]

@taojing2002 can you elaborate on why you are removing that identifiers table? It strikes me as being important, and that your action here does not solve the problem described in your ticket. More info would help me understand. Thanks.

[taojing2002]

@mbjones I was thinking the one function of the identifier table is to figure out the internal docid, then figured out the object file location. This function will be obsoleted when we use the checksum approach in the 3.0.0 release. So I thought we didn't need the table any more. However, I forgot we need the identifier table to keep the list of all identifiers no matter they still exist or haven't been deleted. So I am going to reopen the ticket.

[mbjones]

That's exactly what I was thinking too for why we needed the table still, so I think we are on the same page now. Thanks for clarifying.

[taojing2002]

We can keep it as is.