Community impressions

On Thursday I attended the Leipzig Data week and gave a presentation about our current discussions. Below are some impressions, without claiming a representative nature.

As context, the organizers of the event are long term contributors to the RDF community, as researchers, tool providers and use case owners.

The keynote from Bernd Gardner, AstraZeneca, showed a strong use case, to enable scientists to find information for their research. Bernd said in his presentation that there are a lot of discussions around RDF and PP. And the strengths of using both (for data integration vs. analytical capabilities) are clearly seen.
Bernd said that in AstraZeneca such combinations take place.

Feedback on my session was very heterogenous. Some people said, “more expressivity is good”, but it was hard for me to see the clear use case behind. Others emphasized that even RDF star will lead to issues in the RDF application and tool ecosystem and questioned the need for the burden that RDF star will create, no matter if we go for “one to many” or “many to many”.

Some thoughts on this feedback below.

I understand the reasoning of the group concerned with the effect on the RDF ecosystem, and it triggered my workflow related comments in the call yesterday. With the upcoming update of RDF, we will have already multiplications of potential variants for each workflow step.

Data producer
Intermediary 1
Intermediary n
Data consumers
2 variants: RDF 1.1 vs. RDF 1.2
4 variants: RDF 1.1 vs. RDF 1.2
8 variants: RDF 1.1 vs. RDF 1.2
16 variants: RDF 1.1 vs. RDF 1.2

This is a price one needs to pay for the move to more LPG vs. RDF compatibility.


With the variation of one to many vs. many to many, we get more variants.
Data producer
Intermediary 1
Intermediary n
Data consumers
3 variants: RDF 1.1 vs. RDF 1.2 one to many
9 variants: RDF 1.1 vs. RDF 1.2 one to many
21 variants: RDF 1.1 vs. RDF 1.2 one to many
63 variants: RDF 1.1 vs. RDF 1.2 one to many


Souri pointed out that validation can help to capture the variants. But that does not lower the burden to remodel the data. And with more potential variants, more needs for remodeling will happen.


Best,

Felix