Incorrect `RoW` market shares for freight train transport

[cmutel]

In the dataset market group for transport, freight train (GLO), we get a non-zero PV in the ecoinvent release:

[cmutel]

Ocelot is doing the logical thing here, but this is perhaps not what the reference system expects. The problem is that the definition of RoW in market for transport, freight train excludes the USA (which has its own market). However, the reference system wants to link to transport, freight train, electricity in RoW. However, this is a different RoW, which excludes only Switzerland, China, and Europe without Switzerland. The market linking algorithm only allows links for market suppliers which are contained within the market location. And the RoW for transport, freight train, electricity includes the USA, and so is not contained in the RoW for market for transport, freight train.

[cmutel]

Comment from Ingo Meinshausen (email from 4.2.2019):

RoW is not resolved per dataset. I.e. the linking process does not
know which specific locations are contained in each RoW. With regard to
spatial operations RoW never intersects other geographies.

RoW never intersects anything else, not even another RoW. GLO
contains everything but RoW. But keep in mind that the market supplier
linking works different for RoW DS! Their suppliers are not gathers by
spatial operations but by using suppliers that are not supplying local
markets. If none of those exist the supplier with the largest PV will be
chosen.

Should be functionally similar to doing the "right" thing in most cases, but not here. Happy happy fun times!

[cmutel]

This particular problem was resolved in 6b61152, but some issues around this topic remain.