Meeting minutes
Draft minutes
<kaz> Nov-7
Ege: https://
Ege: Any objections?
… none -> minutes approved
Initial Connection
<EgeKorkan> Initial Connection proposal on Hackmd
Ege: back to the document we talked about yesterday
… cleaned up the document a bit
… propsosal building up on Lucas idea
… what we call "a valid form" is still the same
… we have examples with all the different variations
… connection is basically copy & pasting into connection
McCool: security defintition should come earlier
… we should not repeat the same scheme etc again
Ege: you get both benefits
McCool: I think it is not consistent
… we should use the mechanism we have
… I still think we should not be re-doing the definitions in a different place
… verbosity is not an issue
Kaz: Meta question
… how long do we want to use the time for initial connection
Ege: Depends .. once finished we can switch to use-cases
Kaz: OK. If we'd like to prioritize the Initial Connection topic for today right before Plugfest, that's fine. Just wanted to confirm that.
Luca: Very strong "NO" to keep security definitions out
… duplicates things twice
… security definition is always a connection
… we discussed that lenghy in past
McCool: Question do we ever use security more than once
… combo schemes might cause problems
… complex examples blow up
Luca: Make connection definitions "stackable" would address your concern
… having 2 containers doing the same, I don't like
McCool: Does "stackable" mean sub classing?
… is that what you mean?
Luca: Yes
McCool: I am surprised to see this approach.. Orthogonal might make more sense
Ege: we don't give new names
… let's look at examples
<McCool> (sorry, I was muted, but did say that "security" should be "scheme" if we take a subclass approach. Defaults etc. would also apply.)
Ege: inline version
… 1 protocol, 1 contentType etc and form becomes rather empty
… equivalent example .. just more verbose
… just not inlining it
… example without any re-using
… like the current TD
… next example, without defaults
… what we do not allow is to inline the JSON object instead of the key
McCool: What do you inline.. connection or security?
Ege: string is a reference
McCool: About security
… inlining security is similar idea
… we cannot really inline security
… w.r.t. sub classing vs. delegating
Ege: You are talking about consistency ?
McCool: Yes, having string and object for security
… like anonymous scheme
Ege: Concern about mixing?
McCool: Yes
… not clear and we could have name conflicts
… similar idea for connection inlining
Ege: I would say for parsing this might be annoying
… more if conditions
… anyhow, part for security makes sense
… Luca?
Ege: Showed proposal within Siemens. Liked it so far .. but ask for straightforward parsing
Daniel: Difficult to comment.. I think we should put it into an GH issue with a nicer comparison
Ege: BTW, for Plugfest I will provide a script that converts one to another
McCool: w.r.t. parsing complexity ... use different names
… I think this removes some IFs
Ege: Yes, partly
Koster: Alternative design: make references explicit
… used in SDF
… with key-word ... it makes it easy to parse
… $ref
<Ege showing example to override pre-defined connection in form>
McCool: I think this is different
Koster: Yes, it is an alternate design
Ege: Overriding should be possible, M. Koster?
McCool: Another facet is that it should be easy to understand by people
Luca: About parsing key or object
… is it sort of common
… strategy is to look-up key
… you create dictionary
… I think it is common and straightforward
… w.r.t. getting the names
… forms contain a form ... saying "default" seems wrong. we should come up with a different name.
Ege: Maybe we could say "connection"
<McCool> ntd... I have another meeting starting on the hour
Luca: "inherit" ?
<McCool> re use cases, we can discuss next week in the UC call and/or in the TD call once I have had time to update the PR
Luca: concept of "combo security" makes the whole thing rather complex
… same problem exists already with combo container
… should look at complex examples
Kaz: My impression is that the example is getting complicated more and more like "SVG and MathML within HTML5".
… wonder whether connection definition should be inlinable
… might want to think about defining connectionDefinition separately
Ege: Out-sourcing connection definition might be ideal for some cases but resolving issues pop-up
Ege: I will try to show all the examples in a nicer way
[adjourned]