See also: IRC log
<azaroth> trackbot, start meeting
<azaroth> scribenick: bigbluehat
azaroth: going over the agenda
... there's been a flurry of i18n discussion
... we'll get some updates on testing
... if there are any demos, it'd be great to see them
... is there anything anyone would like to add?
TimCole: I have a couple questions for clarification about model testing
azaroth: k. we'll run with what we have
<azaroth> PROPOSED RESOLUTION: Minutes of the previous call are approved: https://www.w3.org/2016/07/29-annotation-minutes.html
<TimCole> +1
+1
<azaroth> +1
<takeshi> +1
RESOLUTION: Minutes of the previous call are approved: https://www.w3.org/2016/07/29-annotation-minutes.html
<tbdinesh> +1
azaroth: any announcements?
<azaroth> I18n issue: https://github.com/w3c/web-annotation/issues/335
azaroth: we have one major issue
... hopefully people have been keeping up a little
... there was some discussion in the Social Web WG
... about how to resolve their I18N issues
... there was some suggestion that they consider how we solved our I18N
issues
... they disagreed, and have told us we shouldn't do it that way either
... there are three folks from social and one person from Europiana who
feel we need to change our approach
... there are also two subsidiary issues that have come out of it
... one that proposes getting rid of textDirection and instead use
Unicode characters
... and 337 around the carnality of processingLanguage
... We have to address them. the question is how.
... We did talk to the I18N folks, and this approach was OK with them.
... reversing them would be tricky as they've not been marked at risk
... and in my opinion we have not seen information that proves they are
actually not useful
... we could change how we've described them
... but we did go around several times with the I18N folks to come to
the descriptions that we do have
TimCole: one thing that got mentioned, is
the assumption that we are specifying client behaviours as well as the
data structures
... the issue of context arose around formatted language hints
... that an annotation creation agent could add about a resource
... that might be useful
... and that allowing multiple languages listed made it less useful
... so we added processingLanguage and textDirecton to help
... but then we've also stated that if the actual resource states
something different, that these can be considered as overridden
... and these properties are meant to be used as hints
... since we're not specifying client behavior, I don't think we should
put anything heavier in place
... I'd propose some editorial revisions that clarify the intended use,
and clarify that we don't know how they'll be used, but that we wanted
them there as we felt they were useful
azaroth: format and language we can specify
them on the annotation that the client can expect these things, but we
can't guarantee
... maybe the name has changed or whatever else
... this is true with everything on the Web
... the canonical source is authoritative
... and it seems like a general principle
<PaoloCiccarese> +1
azaroth: in Section 2
... saying we're not trying to be authoritative
... you can add information as it's helpful, but if it's contradicted by
the source, then you should believe that source
ivan: I'm not deep into these topics, but we
did have a very long discussion with the real experts
... and with all my respect for the detractors, we've made our decisions
with the input of these experts
... and with that as background, we should not remove them
... unless they I18N folks ask that after all they would rather we
remove them
... as long as the I18N experts maintain their advice, we should keep
things as they are
<tbdinesh> +1
<azaroth> +1
ivan: I am not sure what exactly what the
editorial explanation like comments that we got
... until now the only thing I see as a cursory reading is additional
technical discussion
... and doesn't seem to result in a "please put this content in this
place in this way" sort of change
... and it's really just too late for this sort of editorial usage
commentary to be added to the spec
... So, unless the I18N experts suggest we change them, we should make a
resolution that we intend to keep them based on their input
<ShaneM> +1
<tbdinesh> Unicode indicator will override. so not contradicting anything
ivan: and that it's just too late in the process to add the commentary
<azaroth> PROPOSED RESOLUTION: Unless the i18n group advise otherwise, the WG will not remove textDirection and processingLanguage from the model
<PaoloCiccarese> +1
<azaroth> +1
<tbdinesh> +1
<TimCole> +1
<ivan> +1
+1
<takeshi> +1
<bjdmeest> +1
<ShaneM> +1
PaoloCiccarese: so. I have a question.
... say they don't come back to us about it
... and that in a few months they come back and say that we should
change these?
ivan: this can happen to any feature in the document
<ShaneM> Errata is a way to capture information
ivan: it is a matter of at some point in
time we'll have to figure out an errata mechanism
... if it happens at all, it would be handled as an errata
<tbdinesh> how can it be an errata. its a MAY
azaroth: are their any other thoughts?
RESOLUTION: Unless the i18n group advise otherwise, the WG will not remove textDirection and processingLanguage from the model
azaroth: the other one ivan proposed seem to say if we could come up with something better we would
ivan: yeah. we've mostly already said this via other comments
azaroth: correct, we've said essentially "if anyone has other commentary, we're happy to hear it."
<ShaneM> as long as the change is non-normative!
ivan: right, as long as we've stated that we're here and accepting input
<azaroth> PROPOSED RESOLUTION: WG is happy to improve the text of the textdirection, processinglanguage descriptions with clear input as to what those changes should be
+1
<ivan> +1
<tbdinesh> +1
<TimCole> +1
<azaroth> (and indeed, any other text in any of the documents!)
<azaroth> +1
<bjdmeest> +1
<takeshi> +1
<ShaneM> +1
<PaoloCiccarese> +1
RESOLUTION: WG is happy to improve the text of the textdirection, processinglanguage descriptions with clear input as to what those changes should be
takeshi: I noticed that processingLanguage does not state what values should be used
azaroth: yes. you are 100% correct.
takeshi: to make things administratively
right, let's get that added as a GitHub issue
... so it has a place, and the editors can handle it
... because we need to have documentation for all our changes
<azaroth> PROPOSED RESOLUTION: Clarify that the value of processingLanguage SHOULD be bcp47, as per language.
<TimCole> +1
<azaroth> +1
<bjdmeest> +1
<takeshi> +1
<tbdinesh> +1
<PaoloCiccarese> +1
+1
azaroth: do people think we should invite the I18N folks to the meeting next week?
ivan: probably, we should show public support--again--to say we support their input
RESOLUTION: Clarify that the value of processingLanguage SHOULD be bcp47, as per language.
<azaroth> trackbot, pointer
<trackbot> Sorry, azaroth, I don't understand 'trackbot, pointer'. Please refer to <http://www.w3.org/2005/06/tracker/irc> for help.
azaroth: I will make an issue that says that our content is based on the input of the I18N group
azaroth: one key piece to discuss is testing
the value of properties even if the property is optional
... including format and language
... there is an ever increasing number of media types
... and there's on the order of 8k languages and sub features on those
... that also in theory could be tested with a JSON Schema
... but that seems like massive overkill
... and seems impossible to maintain
... given the ongoing addition to both lists
... so perhaps we can test the shape of the strings
<ShaneM> ^[a-z][a-z][a-z]?
azaroth: using the list of top-level media
types followed by a slash followed by some other characters
... is probably sufficient for media type
TimCole: I agree that if the feature is
optional, we won't try to be perfectly rigorous, but we will try to warn
if the shape is wrong
... that seems better than not testing at all
... the perfect solution would need to use an outside services
... we can continue to test the MUST requirements more rigorously
... the other issue that came up was that if testing the value it's
possible to add some more clarity to what the value should be
<ShaneM> it means we need a separate assertion
TimCole: there were a couple of other things
... there are situations where we reference external vocabularies
... two keys on Annotation:
... audience and accessibility
<azaroth> https://www.w3.org/TR/annotation-model/#intended-audience
<azaroth> (and following section)
TimCole: the audience one should be from the
Schema.org namespace
... these come from the Schema.org classes
<azaroth> ^schema:(.+)$ ?
TimCole: right. as azaroth says you can check that the value starts with `schema`
ShaneM: a follow-up question. Does the spec hard code the prefix?
TimCole: yes.
ShaneM: excellent
TimCole: k. that sounds easy and we don't
need to look up the actual classes
... and then the last question
... we do have a few situations where it "should be exactly 1 but may be
0 or more."
... it seems worth checking these
ShaneM: thinking about. it's not a fail regardless
azaroth: seems like a warning if there's 0
or more than 1
... and fine if there's only 1
ShaneM: we have syntax for this in our
assertion structure
... the tests can still continue even if this check fails
... what we need to do is make sure the message that comes out with the
failure states that it's a suggested feature and not an actual failure
TimCole: could you clarify the "succeed and continue" situation?
ShaneM: "succeed and continue" is not the
result, it's the expected result?
... so if the thing worked, continue. which is what you'd expect it to
do
... I think we should just play with the combinations and see what we
get
TimCole: I'd like a message to come out in
the results which is a warning, but that the process does continue
... and its still marked as valid
ShaneM: unfortunately WPT has no method for outputting a message when things are successful
TimCole: so passing the test is fine, but if
you didn't pass the test, then we could "succeed and continue"
... we'll try that see if that's what we need
ShaneM: right. that'd be the approach I'd suggest
TimCole: we also have definitional schemas
and other schemas that can be used as test scripts
... we should be able to try and run some test scripts next week
... and we'll be working on section 4 and section 5
... the goal being to have all the schemas built next week
... and perhaps we could get help running the test scripts
... if I provide some draft test scripts
ShaneM: I'm happy to do that. the test dev environment is actually up and running
<ShaneM> testdev.spec-ops.io:8000/tools/runner/index.html
ShaneM: and stays up to date with the repo
TimCole: k. we'll try that
... the tests could compare a series of assertions
... I'd like to put something together that puts all the optionals
together
ShaneM: I think that's fine. as long as
there's an annotation that matches those criteria
... there's just no way to do it with one huge schema, because there's
no way one annotation would exercise all those tests
TimCole: great. we're hoping to put stuff
together with some of these libraries
... perhaps in another language besides JS
<ShaneM> https://www.npmjs.com/package/language-tags is a JS package to validate language tags btw
azaroth: happy to help with Python for one
TimCole: if nothing else is needed here, let's move on to protocol testing
<TimCole> scribenick: TimCole
bigbluehat: protocol multiple headers was a bit of a red herring
<ShaneM> https://www.npmjs.com/package/media-type JS library
bigbluehat: have a stack of PR to bring protocol more in line with LDP
ivan: are they all editorial?
bigbluehat: maybe not, e.g., prefer
headers...
... include parameter in 2 headers, the latter should override
... have rewritten the spec language to make clear how multiple headers
work together
ivan: a little lost in details
bigbluehat: so this is the section that I've
proposed be revised
... includes how LDP specifies the link headers should look
ivan: this seems making more precise, not a technical change?
<tbdinesh> yes there is
azaroth: the CR doesn't mention multiple headers, so perhaps this is a clarification
<ShaneM> https://www.w3.org/TR/annotation-protocol/
bigbluehat: we did add an 'example 4', was
not in CR
... so this is editorial
ivan: any change (even editorial) has to be fully documented and determined to be editorial
azaroth: for the PR for this protocol issue, can we go ahead and merge after update section is updated
bigbluehat: Note, the PR includes a few
other clarifications
... will add something to the changes section and then azaroth can merge
azaroth: how close to protocol testing implementation?
bigbluehat: updated to do include headers
properly
... researching how to get into WPT
<bigbluehat> https://github.com/Spec-Ops/web-platform-tests/pull/3
ShaneM: at point where we want to request merge into WPT (once mergeable)
bigbluehat: yes. and once merged should be fully compliant annotation server
<tbdinesh> is there an annotations "dump" somewhere that can be used for protocol testing?
azaroth: anything further on protocol?
<Zakim> ShaneM, you wanted to ask a question about model testing (before we adjourn)
ShaneM: found js tests for media types and
languages
... ajv supports this
... does not rely on external services
... ajv implementation does not rely on external services
azaroth: would tie implementation to ajv
libraries
... would not be able to use in python
... if possible to have a branch?
TimCole: wait and do later, but not slow down progress on implementation testing