Meeting minutes
<Laurent Le Meur> The draft spec is at https://
<Laurent Le Meur> Thanks to Ivan for the editorial cleaning.
<Laurent Le Meur> issue #2835: the creator on an annotation can be a software agent.
<Laurent Le Meur> -> Gautier
<Laurent Le Meur> #2850: Consider multiple values for Annotation tags
<Laurent Le Meur> -> Readwise, Zotero
<Laurent Le Meur> #2851 to 2857 from Ivan Herman, yesterday
Laurent Le Meur: Looking at spec. There are no issues in the text
… we do have some open issues in the tracker we need to discuss
issue #2835: w3c/epub-specs#2835
Laurent Le Meur: This is a request to add a software agent as creator of an annotation
Gautier Chomel: We already discussed with DAISY, there use cases where we might want these displayed as a set
… and you may have tools that generate annotations, e.g a tool from ACE
… So it would be good show this to the user
Laurent Le Meur: At the set level we already have an entry for tool
… so in this case ACE would be the tool and the creator
… Anyone against?
crickets: no noise
Laurent Le Meur: Ok, I will add it
w3c/epub-specs#2850
Laurent Le Meur: This is a request for a list of tags
… Readwise specifically asked this
… they also want the date the tag was made, but that isn't critical. A simple list would be fine
… are there other implementors on the call who are interested?
Abe Jellinek: We are interested
Laurent Le Meur: What do you refer to it as?
Abe Jellinek: Just tags
Hadrien Gardeur: I think if we support tags then we have to support multiple
… otherwise we risk ad hoc systems
Laurent Le Meur: I think we have to support tags, so therefore we need multiple
… the group seems to be ok with that
… And it yesterday Ivan added several issues, but I haven't had a chance to review them
Vocabularies
Ivan Herman: This is json-ld, so it is by definiton RDF
… so we have to precise and formal
… This sounds scary (making it a proper vocabulary), but it isn't so bad
… I have a tool that takes a yaml file for the vocab, and it generates a vocab description
… I generated an annotation vocabulary uses the current doc and all the terms in it
… the classes are pretty good, but the properties are messy
… for instance we may need to create sub properties, and a few other things
… demo of generated vocab
… Shows classes, domains, etc
… I won't go into the details, but it also makes turtle and json-ld versions
… question is how far do we want to go with this?
… the problem is some changes constrain the general term
… e.g. creator must be less than the original. So you can't really do this in rdf, it is just in the text
… we have to decide if we want to go over all the details or not
… we also need to decide on how to handle it editorial (separate doc, appendix, etc)
… The issues are ones that came up during this process
… none are really complicated
… we can just take them individually, they are independent of the actual work
Laurent Le Meur: Ok, that's great! The audience for this will be small
… when you say should we be general or go into details
… I think we should consider this like a schema
… you can't say everything in the schema, we should leave that to the text of the spec
Ivan Herman: Yes, and the the way it works is it makes a link back to the original core spec
… so it doesn't repeat the text of the core spec
… we really need a second pair of eyes to review the results
Laurent Le Meur: Should we discuss any of these issues today, or wait for the next call?
Ivan Herman: [looking] there are some editorial things
… There is one interesting one
… You introduced language and direction as terms, which comes from web annotations
… that was not done with an earlier json-ld
… but json-ld has added these terms
… so I think we should just rely on those
… an other one: in annotation set there is an items property. Are these sorted?
Laurent Le Meur: For EDRlab it is unordered
… there is no sequence, the order is meaningless
Ivan Herman: I agree, but it needs to be clarified in the spec
… another thing: at the moment, for two classes, the content reference context is a required property. I think that is wrong
… there is a serialization section, there should be a reference there to the context that should be used
Laurent Le Meur: There is an issue for the rdf (json-ld) when there are 2 urls
… can we just have 1?
Ivan Herman: It is possible to import a context file, but security people don't like it
Laurent Le Meur: I am really thinking of the developers
Ivan Herman: The others are mostly editorial
Zotero vs the draft
Laurent Le Meur: ajellinek did you have a chance to look at the spec?
Laurent Le Meur: Could you provide any insights?
Abe Jellinek: Our comments are largely techinical
… in terms of what we have so far I like it
… some uris are tricky in our data model, but those are all solvable
Laurent Le Meur: So with readwise and zotoro relatively happy, we can probably get to 3 implementations
… next in January will be selectors
… the question will be which to take from w3c annotation model? And which to add?
… and after that, which are mandatory for interop?
… if nothing is mandatory we face non overlapping sets of selectors from developers
Official publications
Ivan Herman: Publishing the 2 docs as official w3c docs would be good
… we don't need a final version, this just needs to FPWD to help get reactions from others
… The current draft seems good for FPWD
… so let's do it ASAP
Laurent Le Meur: We have to decide whether to do it before or after selectors
Ivan Herman: FPWD doesn't have to be implementable, it is just a direction
… maybe just put in the fragment selector as an example
… that is good enough
Laurent Le Meur: Then we can do that in January
… how do we add these discussions to the issues?
Ivan Herman: it is auto, it will take it from the topic setting