[BioSQL-l] Remote locations

[Hilmar Lapp]

To identify the sequence for remote locations there is now a table 
Remote_Seqfeature_Name with two attributes accession and version, 
and an identifying FK to Seqfeature_Location.

I propose to replace this with a nullable FK in SeqFeature_Location 
to DBXRef. At the same, I propose to add an attribute version to 
DBxref.

Here's why:
1) The current design takes only locations on seqs in the same 
namespace (as it is implicit from 
seqfeature_location->seqfeature->bioentry).

2) My proposal would treat the remote seq more as what it is (a 
dbxref). At least IMHO.

3) The nullable FK to dbxref would be compatible with a nullable FK 
to bioentry (if you, as I did in a local version, model dbxrefs as a 
skeleton view on bioentries).

4) If necessary to improve performance, a nullable FK can be be 
joined in in one hit (outer join), unlike an identifiable child.

Does anyone have objections? Also, comments/opinions appreciated. As 
before, this change would only happen on the main trunk.

	-hilmar
--
-------------------------------------------------------------
Hilmar Lapp                            email: lapp at gnf.org
GNF, San Diego, Ca. 92121              phone: +1-858-812-1757
-------------------------------------------------------------