<kaz> scribenick: andrea_cimmino
christian from MongoDB is present
meeting starts with christian presenting himself
<kaz> (Christian has read the W3C Patent Policy and understands the RF commitment)
McCool: minutes from last meeting are reviewed
<inserted> August-3
McCool: PR for anonymous TD was
created, as stated in the minutes
... remind McCool to create issue for the spec issues/43
... remind McCool during MongoDB discussion to create an
idempotent issue
... minutes proposed for publication with no further
comments
minutes reviewed and publish, changing topic now to MongoDB
McCool: Does MongoDB supports JSON-LD? and XPath?
Christian: it supports
Json-LD
... regarding XPath, MongoDB supports a similar query
language
McCool: it would be good to have
the spec of such language
... the place to collaborate is to have a second implementation
based on MongoDB
Farshid: current implementation
is not based on MongoDB
... instead the implementation uses LevelDB
Christian: how does LevelDB is used?
Farshid: the key is the ID and the value the TD
McCool: the TDD have an API to
register, and then someone can filter the TD. TDs have an id
field, which is optional, therefore sometimes the ids must have
a created on the fly id
... there could be other meta-data, also not included.
Therefore we rely on a wrapper object to support such
information
... ids can be updated, they are not immutable
... the general idea is that authorized devices can receive
notifications, so basically, when the TD changes de id it has
to be deleted first, re-created, and then notify those
devices
... when created on the fly, maybe, ids should not be changed
if TD is updated
... bottom line, the draft needs to include how to deal with
this stuff
... implementation time should be until January 2022
... candidate recommendation July 2021 (CR)
... final for December 2021
... ideally, July 2021 we should have two implementations that
satisfy the requirements
<kaz> (Implementation report plan by CR transition; Implementation Report with concrete results by PR transition)
McCool: implementations do not need to be Open Source, but they must pass the tests
farshid: our implementation is
based on the requirements of a project, from which we got
involved in this initiative
... there are things outside the specifications, like how the
data is stored, and instead, the specification details the
APIs
... in this way someone can implement internally the spec with
the desired technologies, but still implement the same spec
McCool describes how the current draft is structured
McCool: the draft does not
specifies the internal functioning
... one of the issues are the type of search supported: XPath,
Json Path, and SPARQL
... we will need a SPARQL-based implementation if SPARQL is
included
... Xpath and Json path are still under discussion
<kaz> (Kaz's note: basic expectation is all the features including optional features should have more than one implementation; but there should be multiple implementations given SPARQL is also a W3C Recommendation.)
McCool: clarification, the
project of Farshid was originally LinkSmart, from which other
projects were derived and this component enhanced
... having an open SPARQL can be dangerous, therefore, it will
be used for advanced implementation
... the baseline is Json Path and XPath
... in general, regarding the architecture, privacy is a
concern
... since distributing data can be used to infer information
and also track people
... in the discovery there will be two steps: introducing first
(only informative), and then the exploration mechanisms
... even if christian does not implement a version, your contribution and experience
will be very valuable
... for instance, in the idempotent issue
Christian: the problem here is
how to identify anonymous TD
... a hash could be calculated, but the order of the keys is
important
... the full td should be used for the hash, so two TD are
identical only if their whole content is the same
McCool: here the thing is that the same device could create different TDs that are the same, exactly the same content
Farshid: some UC may include this
beahviour
... for these cases maybe a short life span may help
... two devices are identical, and publishing the same TD, and
only their payloads are different
... these, once registered, get a unique ID. But this ID is
unknown during registration. Everything should work fine
McCool: maybe we should register
a special TD, get the ID, and then register the right TD with
the ID
... let's further discuss this
... let's create an issue and discuss (assigned to farshid)
<FarshidT> Issue to discuss how to handle anonymous TD and avoid duplicates: https://github.com/w3c/wot-discovery/issues/48
<kaz> PR 38
discussion about last befrancis issue, farshid already replied
Farshid: the comments are outside
the scope of directory
... like, having a TDD that automatically creates TDs from
adapters
McCool: maybe we could approach
this creating micro-services that perform different tasks
... nevertheless, McCool agrees with Farshid
McCool replies in the PR
McCool proposes merging the PR, since no objections are done, PR is merged
scribenick: kaz
<kaz> PR 47
<kaz> diff
<kaz> changes
McCool: (captures several issues as comments)
<Zakim> kaz, you wanted to suggest we use <h2> with <section> for ReSpec. Please see also: https://github.com/w3c/respec/wiki/section
<kaz> (<h5> surrounded by <section> and </section> is OK, though)
[adjourned]