Exchange format

I like Glen's practical xml format for the variation data.  The version made from the workshop model was just an attempt to directly capture info from the model and it was getting too... complex.

There is a need to annotate variants based on reference sequences, which can be different splice variants. Just wondering would it make sense to handle that by optional refseq (or it can be more general dbxref) attribute like this:

<variant>

		<id>UBE3A_00001</id>
		<exon refseq="NMXXYYZZ"> 08</exon>
		<dna_change refseq="XYZ">c.3_16del14</dna_change>
		<rna_change refseq="ABXY"/>
		<protein_change>Frame shift (predicted)</protein_change>
		<restriction_site/>

Should not be a problem if there is a possibility to annotated elements using different refseqs. ...Actually it is good to have the freedom I think. 

Note also that there can multiple exon and dna/rna change elements in the variant-element. 

In this way we can also include legacy-annotations as well