RDF and the real world [tm]


#1

Every now and then we have the “nobody is using RDF”, “RDF does not work because…” or “RDF is complicated posts” going on. Most of the time, but not always, by people who might have done some work with RDF but no longer do so. You can find that on Twitter, Reddit, blog posts, Medium, etc.

The RDF research community itself is also famous for talking a lot, I remember some pointless discussions at the W3C Graph Workshop in Berlin in 2019 or the discussions that pop up at semantic-web@w3c.org on a regular base. I will not link to them because they do not add any value to the part of the RDF community that mainly wants to solve real-world problems.

I always say the same: RDF is not the problem, tooling is. I call bullshit on “RDF is too hard” or “SPARQL is difficult to learn” every time someone claims it. I’ve done amazing stuff with RDF and SPARQL and I teach it to newbies on a regular base and most of them get it fast.

But as long as these discussions pop up, we still have work to do. Apparently we cannot transport the benefits of RDF and its stack to others. This is a meta-post to collect some things I would like to talk about:

  • RDF is hard, based on a tweet by Melvin. Which IMO shows why RDF is the stack to do stuff that last longer than the developer(s) think. I gave some ideas on why and will elaborate on them.
  • RDF standards vs implementations in code, based on a post by my colleague Tom. As Dan mentioned, this is more a question of Open Source sustainability in general than RDF itself but in our community it seems especially hard.
  • Re-inventing RDF. This happens on a regular base and I think this deserves a post in its own. The latest example at the time writing is this initial post by LinkedIn, we will see in which direction it will go once they publish the second article.
  • The success of others in the graph-domain, namely Neo4J and others in the property graph world.
  • Stuff we neglected so far. I have tons of idea in my head and we try to get there at zazuko.com but I’m not alone. There are some remarks from Miel here for example. I recently also created this Hackmd document to distract from the pointless blank-node discussion.

I thought about doing that in a blog first but I tend to overthink and then never publish stuff if I do it in my blog. So let’s try this forum instead.


#2

Correct!

There is nothing wrong with RDF bar terrible marketing communications and a chronic unawareness of existing productivity tools.

RDF is supposed to empower everyone, but it morphed very quickly into a “you must be a programmer” realm.

Corrective Action?
We need a directory of RDF Productivity Tools that’s published using Linked Data principles.

Related


#3

Trying to “empower everyone” might be a too ambitious of a goal at this point still if we’re not doing such a great job empowering developers first

Having published that linked post I sparked some interesting discussions on Twitter which made me realize some things.

What @kidehen calls “terrible marketing” I find the result of RDF being developed with a flawed mindset. In the example of R2RML, I find it a very well defined spec but it should not be the goal itself. To be complete it should have been from the start accompanied by practical use cases and case studies showing others what they gain by using R2RML and similar tech to transition from SQL to Knowledge Graph. Academia has been blamed many times (ie. research being the ultimate goal).

Corrective Action?

I cannot (fully) agree with the directory of tools, especially if required to be published in “Linked Data principles”. Unless minimal in scope, it sounds like groundwork for endless ontological discussions and another ivory tower.

Instead I would like to see more resources which can resonate with those uninitiated and what they might have learned:

  1. How Knowledge Graphs solve software problems better than other technologies
  2. How KG can be used together with well-established modeling techniques
  3. How KG interoperate with rather than replace existing implementation methods
  4. And yes, maybe using Knowledge Graph instead of RDF/SemWeb as the overarching brand?

#4

Ironically, even if the LinkedIn article introduces just an RDF-inspired property graph, it might be a very effective steppingstone for RDF advocates.

With great outreach LinkedIn has it seems like a great opportunity to ride the wave to come and show people that similar technologies have long existed right under everyone’s noses.