A mandatory identifier and reference element or attribute is probably needed

[csc-jm]

The alias attribute is currently set as an optional attribute of all objects. To my knowledge all the BP metadata objects are connected and the alias attribute is currently used for cross referencing these objects with one another. Hence, the alias attribute should probably be required or else identifying objects with another is not possible. Either that or create another attribute/element for this referencing.

[erikogabrielsson]

I think that the alias is optional as it is only needed when submitting. Once submitted the object should get a accession number from the repository. But this is just from how I think SRA is using the reference... And if that is the case then there should be a choice between having an alias or an accession number (with restriction that at least one of them should be set) instead of both of them being optional.

[erikogabrielsson]

See for example https://ega-archive.org/submission/sequence/programmatic_submissions/working_xml

[csc-jm]

Yes, we for example create our own accession ID's for each metadata object in our metadata-submitter. However, that happens upon submission. To verify the linkage between for example an image and its dataset, there should be some mandatory attribute or element in each object, which includes this alias/accession so the linkage validation can be done before the submission can go through.

[erikogabrielsson]

We use SRA RefObjectType to reference between objects. Looking at this type, it might be incorrect that we (as in the bigpicture metadata interface implementation) couple refname in the RefObjectType with alias in the ObjectType, and that we instead should use the IDENTIFIERS present in both.

Or maybe we should make our own types for identifiers and references?

[csc-jm]

Using IDENTIFIERS could work just as well but I think EGA does it exactly like this with alias and refname

refname attribute: when an object references another by its alias, the alias goes into the refname attribute. For example, if a sample has the alias "sample1", and an experiment uses this sample, then the EXPERIMENT/SAMPLE/refname should be "sample1".

Even then IDENTIFIERS and its child elements are also optional. Also refname attribute is currently optional. I don't have strong opinions nor expertise on the proper metadata schema conventions, but if we don't feel like altering the original SRA schemas to accommodate the possibility of validating each metadata objects reference, then creating own types for these mandatory identifiers/references could be good.