- From: Alexandre Bertails <bertails@w3.org>
- Date: Mon, 20 Jan 2014 14:35:51 -0500
- To: John Arwe <johnarwe@us.ibm.com>, public-ldp-wg@w3.org
On 01/20/2014 08:39 AM, John Arwe wrote: >> With your proposal, LDP will depend normatively on this, so we'll have >> to hope that these dependencies will become stable soon enough for LDP >> to go to REC. > > Speaking as a certified lead-lined tinfoil hat wearer, the easy solution > is to make it At Risk and make the call at CR. If we dump it at CR, then > clients get all triples in 1.0, this probably lands on the ".next" list, > and anyone who cares can implement the draft as an extension. Fin. Sounds like a plan. >>> from the client) would normally return them. LDP could try Must'ing > that, >>> at the risk of eliciting more comments along the lines of Mark Baker's >>> (and perhaps Erik W, and IETF generally) >> [snip: Alexandre's insertion] >>> that we're intruding on base >>> specs by over-specifying how servers form representations most > appropriate >>> to the client according to WebArch. >> >> Can you say why you think this is "over-specifying"? > > I'm not saying that I think it is, I was anticipating others' likely > responses based on past experience discussing their comments (and their > reasoning that led to same) with them. My perception is that they are > strong advocates for a point of view that gives the server huge amounts of > freedom in determining what the "right" response is for a given situation. > It's that position that leads to conclusions like "200 + a subset of the > state with links to more of it is fine" (how to respond to a GET when the > server decides to page the result). While not everyone agreed with that > specific conclusion, it's a point of view that I see coming through in > many IETF specs. "over-specifying" was my way of summarizing that > position, is all. I understand. But we're defining a protocol so I personally won't say that this is "over-specifying". >> 1. Does this replace 5.9.1 altogether? Do we still expose non-member >> and non-containment resources? > > If we choose to adopt this mechanism to ALSO cover the > non-member-properties case, I think exposing them is server choice. Any > minimalists would probably then conclude "remove all mention of them > [non-member-properties , non-containment resources] from the spec - it's > complex enough already". I would have sympathy for that position, > personally. So would I. I honestly don't think this is crucial to LDP 1.0 given the Prefer approach and I would support removing it for the sake of simplicity. And if this is really important, that feature could easily be re-added in LDP Next anyway. I think it'd be easy for Arnaud to make the group vote on that PROPOSAL. I won't be surprised if we have everybody ok with "remove all mention of them [non-member-properties , non-containment resources] from the spec". And if somebody is strongly opposed, then let's just keep it. >> 2. The spec does not prevent inlining (nor encourage it) and it _can_ >> be useful for a hashless-ContainerResource to inline the containment >> triples from the related containers. I would then expect those >> "normally" inlined containment triples not to be returned if the >> client prefers to omit them. If you agree, do you think we should have >> a remark that mentions inlining in this proposal? > > I don't think the spec even _contains_ the _word_ inlining at this point, > given that we took a decision some time ago to remove it as an At Risk > feature based on LC1 feedback (IIRC *I* removed it, and normally I'd > search before declaring 'done'). I have not looked back at it after that, > I have enough trouble staying current with the LDP email traffic. Given > recent progress on alternatives like TriG that are immune to the > provenance problems that Henry alerted us to, those would be the first > thing I'd look at personally. Ok that's fine. TriG is nice but not necessary in my opinion. If I GET <http://example.com/foo>, I know that as an Linked Data client, I should consider as authoritative only the statements about <http://example.com/foo> itself or the derived hash-URIs as <http://example.com/foo#bar> or <#bar>. I would consider anything else as being inlined. Maybe that could go into Good Practices. Alexandre. > > > Best Regards, John > > Voice US 845-435-9470 BluePages > Tivoli OSLC Lead - Show me the Scenario >
Received on Monday, 20 January 2014 19:35:58 UTC