- From: Antoine Zimmermann <antoine.zimmermann@emse.fr>
- Date: Fri, 24 Jul 2020 21:34:35 +0200
- To: Hugh Glaser <hugh@glasers.org>
- Cc: Eric Prud'hommeaux <eric@w3.org>, Semantic Web <semantic-web@w3.org>, Maxime Lefrançois <maxime.lefrancois@emse.fr>
Hugh, Le 24/07/2020 à 20:53, Hugh Glaser a écrit : > Thanks, all very helpful. > I just have one disagreement, which is not necessarily important to your argument (sorry!), but I don't want to let through without comment. > >> On 24 Jul 2020, at 19:23, Antoine Zimmermann <antoine.zimmermann@emse.fr> wrote: >> >> <snip> >> It truly is the spirit of RDF datatypes to have to parse the lexical form to understand what a literal means. >> <snip> >> --AZ > > I really don't agree with this, and this may be where some trouble arises. > The "spirit of RDF" to me says that the whole point of literals is that is where meaning bottoms out - if you want meaning, it goes in the RDF. > Viewing literals as having structured meaning is what gets people into the trouble of equating names with people, etc.. > And indeed that a literal could be "10 inches" instead of "10". > > I see that what is being proposed around UCUM as very interesting, and moving away from that spirit, but personally I think it is wrong to say that there is no change being proposed. > At an extreme, this would mean having RDF graphs with a single triple that has a huge and complex N3 document in its single literal, with a full RDF semantics. It's good that you mention this because there has not been any argument about why put something in a literal and why not. It is obvious that putting everything in a literal, as an rdf:Turtle datatype or something like that, seems completely silly. What literals should be used for is to convey information about something that can be represented exactly in digital form. So, for example, if I want to tell you about the date 24th July 2020, in Central Europe Summer Time, I can represent it exactly as an xsd:date. The digital information tells you *exactly* what I'm talking about. However, if I want to tell you something about France, I cannot give you a digital representation of Antoine Zimmermann. It's always going to be partial, imperfect, and for sure not referring to the physical entity that consists of my body and mind. A physical quantity like the speed of light can be conveyed exactly, perfectly, in digital form. If I write: "299792458 m/s" is exactly what I'm talking about. There is no missing information. Now, there is some caveat that needs to be expressed: physical quantities are not the same as measurements. If I want to say what the length of my table in my living room, I cannot convey this information exactly. I cannot say that it is exactly "2.5 m". However, I can say that my measurement, however imprecise it is, gives a measure of "2.5 m", which is an exact physical quantity. I still need some extra information to specify the precision: <measurement> :value "2.5 m"^^cdt:ucum; :precision "1 %"^^cdt:ucum . Moving things out of a graph into a literal does not mean that RDF loses its power and that you need not express relations between entities with identifiers, predicates, and all that. --AZ > > Where does it stop? >
Received on Friday, 24 July 2020 19:34:51 UTC