Hi Martynas,

Thank youbthatbs a very interesting perspective.

Your formulation:

Webpage = Transformation(Projection(Dataset))
Webpage = XSLT(SPARQL(RDFDataset))

really resonates with the architectural direction I was trying to describe.
Itbs a very clean way of expressing the idea that many web applications can
be understood as selecting relevant data from a dataset and then transforming
it into a presentation.

I also appreciate your point that a large class of web applications can be
viewed as custom UX over domain-specific CRUD APIs. That aligns closely with
the notion in the paper that much of what we currently implement imperatively
is fundamentally data selection and transformation, and could be expressed
declaratively when the data model is well-defined.

Whatbs especially interesting is that your example shows the same
architectural pattern emerging with a different underlying data modelbRDF
graphs queried with SPARQLbrather than tree-structured XML, even though
XML-based technologies such as XSLT still fit naturally into the
transformation stage. That reinforces the idea that the core issue is not XML
vs. non-XML, but whether the application is organized around declarative data
flow or imperative orchestration.

Ibd be very interested in reading your posts on this. It seems like a
natural extension of the ideas in the paper, especially in terms of
generalizing the projection + transformation model across different data
paradigms.

Thanks again for sharing thisbitbs a great contribution to the discussion.

Best,
Roger