- From: Rob Sanderson via GitHub <sysbot+gh@w3.org>
- Date: Wed, 16 Sep 2015 02:31:49 +0000
- To: public-annotation@w3.org
The issue is a conflation of model and serialization. The desire is to have a "simple" JSON serialization, and that is driving the model rather than aiming for a coherent and consistent model, and trying to fine tune the serialization over top of it. So we end up with overloaded properties like hasBody causing inconsistencies like being able to have a URI with a string for the target being a URI, but not for the body where a string is just a string. We discussed two properties but, if I recall correctly, the serialization of "body": "comment" was the preferred solution. I agree, personally, that two properties would be cleaner if we must have literal bodies at all. Then systems that don't want to use them but instead have a clean and reliable structure can just use the embedded content pattern. As for language, having one pattern is better than two, and you can't have both language and format on a literal. Hence the restrictions on literal bodies to have neither, especially as in JSON-LD the structure would be *identical* to the embedded method. -- GitHub Notif of comment by azaroth42 See https://github.com/w3c/web-annotation/issues/79#issuecomment-140607287
Received on Wednesday, 16 September 2015 02:31:51 UTC