- From: Andy Seaborne <andy@apache.org>
- Date: Mon, 4 Jan 2021 10:59:01 +0000
- To: public-rdf-star@w3.org
On 04/01/2021 08:20, Miel Vander Sande wrote: > Hi Laura, > > I'm glad you bring this up... I was thinking the same thing, but I > refrained from asking to not complicate the consensus ;) > However, note that the outcome of the current effort is not a > recommendation, so syntax can still be changed when going towards a > standard. > > I dig the <<>> syntax, it is in line with the URI notation, but the {| > |} (they are next to eachother on my keyboard btw ;)) feels unintuitive > compared to something like <! >. The { } make me think of graphs That is a downside, but the same argument applies to "[" and "(" for blank nodes and lists. [[ ]] was originally suggested but [[ is already legal Turtle. Analysis: https://lists.w3.org/Archives/Public/public-rdf-star/2020Aug/0043.html and the discussion is referenced in issue 9. > and > that might have been the intention. So to prevent going into this > uninformed (https://github.com/w3c/rdf-star/issues/9 > <https://github.com/w3c/rdf-star/issues/9> doesn't mention the > notation's origins), perhaps Pierre-Antoine or someone else could > briefly explain the rationale. I would definitely consider this an issue > of low priority. A few inputs: 1 * It should be the same syntax for SPARQL as well. 2 * It should work in Trig 3 * It should not more complex parsing technology Turtle and SPARQL are both one token lookahead grammars, without context sensitivity tokenizing (LL(1) and LALR(1) and recursive descent). > Laura, about your [1] proposal: I think this was discussed in other mail > threads. There's the thread by Pat Hayes and Ivan Herman talking about > reference by hash. Lionel Medini also proposed a |:bob :age 42 @{ > :source <http://example.org/~bob/ <http://example.org/~bob/>> > }.| syntax. Bu I believe the plan is to keep annotations and triples > together while staying within the triples model. Yes, that is the key point for annotation syntax. As it is syntactic sugar, you can always write with <<>> :alice :knows :bob . << :alice :knows :bob >> ex:since 1980 . c.f. blank node labels if a triple reference style is used. Andy > > best > > > Op zo 3 jan. 2021 om 09:02 schreef Laura Morales <lauretas@mail.com > <mailto:lauretas@mail.com>>: > > Hello, > since the spec is still WIP and you are welcoming comments, I would > like to suggest to change the symbol {| > The main reason is that I find it very ugly and in stark contrast > with the simplicity and user-friendliness of Turtle. The two symbols > are also on the opposite sides of the keyboard and require quite > some effort to type (at least for ISO keebs), but this is only a > secondary reason; much less of an issue than the first one. I don't > find << >> particularly nice too, but it's completely bearable and I > don't really have much problems with it. But {| |} is just... too > much, I think. > I understand that the symbol must work both for Turtle and SPARQL, > and the list of available characters combinations is limited because > of this fact. So I'm not sure what a better replacement could be, if > a new keyword, or a different 1-char symbol, or a better 2-char > symbol such as {{ [[ (( -> => etc. Can << >> be reused maybe? What > are the use cases for using << >> as an object of another triple? > Maybe << >> as a subject could stand for non-assertion triples, > whereas << >> used as an object could stand for annotation (instead > of {| |}). Even a reference system like this would be better imo: > > :alice :knows :bob . [1] > ...other turtle ... > [1] ex:since 1980 . >
Received on Monday, 4 January 2021 10:59:15 UTC