Re: Proposed constraint

On 2024-01-04 00:50, Nathan Rixham wrote:
> I would like to propose that any WebID specification authored by this group
> must provide a testable way to determine that a WebID denotes an Agent.

It goes without saying that testing is important and encouraged in W3C, 
and I support it for this CG - as I've done in Solid CG: 
https://solidproject.org/ED/qa - where possible.

I'd add implementation experience, i.e., implementation reports to the 
proposed constraints, or at the very least commitments to implement. We 
can operate on honour as opposed to calling the implementation report 
police if someone doesn't follow through.

Let me share a snippet from a comment ( 
https://github.com/w3c/WebID/issues/17#issuecomment-1877081434 ) that 
goes a bit further on testability:

 >"Testing" does not strictly entail automated scripting and 
verification. **In fact**, [W3C EARL](Evaluation and Report Language 
(EARL) 1.0 Schema), which is one of the key outputs of the [W3C QA 
Activity](https://www.w3.org/QA/Activity), provides a set of [Test 
Modes](https://www.w3.org/TR/EARL10-Schema/#TestMode), including: 
"automatic", "manual", "semiAuto", "undisclosed", "unknownMode".

 >So, conformance in a specification can be written in different ways 
towards what would qualify as a product to be interoperable. In 
practise, not every notion or requirement is necessarily tested or 
applicable or only testable with a single test case. That said, of 
course being testable strengthens the legitimacy of implementation's 
conformance through reports. But even then EARL for example acknowledges 
possible [Outcomes 
Value](https://www.w3.org/TR/EARL10-Schema/#OutcomeValue) such as 
"cantTell", "inapplicable", "untested".

 >Certain notions or intentions of a specification may not be practical 
to test. Take for example the notion of [URI 
Ownership](https://www.w3.org/TR/webarch/#uri-ownership) as per Web 
Architecture. If the owner of a URI says that it is allocated to 
represent today's weather, or an agent's profile (document) - and that 
can be stated in many ways - it does not entail that there must be 
specific wording/statement in a given representation with a specific 
testing. If there were to be the case, it would be generally infeasible 
or impractical to test. That's just an example. When I say 
https://csarven.ca/#i is my WebID, it is so, first and foremost, because 
I said so (even just with this sentence). As the URI owner ( one of the 
intended audience: https://github.com/w3c/WebID/pull/29 ), I also have 
the responsibility to manage its representations. To quote a AWWW 
principle: "Reference does not imply dereference". So, when the WebID 
spec defines WebID (the identifier) using an existing URI scheme (HTTP), 
there are limits to what can be tested. (It certainly does not entail 
coming up with several specs.)

 >I've shared quite a bit of considerations in 
https://github.com/w3c/WebID/issues/21#issuecomment-1875278872 
essentially about expressing conformance requirements and interoperable 
product classes, categories of the specification, and weighing actual 
complexity / fragmentation of alternatives.

-Sarven
https://csarven.ca/#i

Received on Thursday, 4 January 2024 13:29:06 UTC