On Sun, 16 Aug 2020 at 03:31, Gregg Kellogg <gregg@greggkellogg.net> wrote: > In RubyRDF, I have indeed implemented both modes, but my feeling is that > we should restrict an eventual spec to SA mode. Looking at JSON-LD, where > we have a preliminary note for JSON-LD* [1], implementing PG mode for > expansion and compaction is impractical, and the transformations really > should be symmetric. > > PG mode could be considered a simple entailment of SA, which should be > described, but I’d say that the base spec should be restricted to SA. > Kinda agreed that PG mode seems both less flexible and more problematic to implement in a performant/scalable manner. One example of less flexible being that PG mode currently does not allow assertion of an RDF* annotation in separate named graph from the one that contains the ground fact. I also am very curious how the various existing PG implementations handle retraction. I'm fine with a spec describing both approaches, though I'd prefer SA as the default and PG as an "optional extra" (perhaps like you say, an entailment extension). We do need the flexibility that SA offers available for our users, though. Cheers, Jeen -- *Dr Jeen Broekstra* *principal software engineer* jb@metaphacts.com www.metaphacts.com [image: htps://www.metaphacts.com/] <https://www.metaphacts.com/>
Received on Tuesday, 18 August 2020 05:58:28 UTC