ISSUE-68: Domain of HTTP links with rel=provenance
http-link-domain
Domain of HTTP links with rel=provenance
- State:
- CLOSED
- Product:
- Accessing and Querying Provenance
- Raised by:
- Olaf Hartig
- Opened on:
- 2011-08-02
- Description:
- From Section 3.1 it is not clear what the semantics of the HTTP link with rel="provenance" is. More precisely, what is the domain of such a link?
Notice, RFC 5988, Section 5.2 [1], defines a default for the domain (called "context of the link" in RFC 5988) as follow:
"""
By default, the context of a link conveyed
in the Link header field is the IRI of the
requested resource.
"""
Was it the intention that we use this default for links with rel="provenance"?
If so, I propose to make that explicit in Section 3.1.
However, I would argue against using this default for links with rel="provenance". Instead, I propose to define the domain (i.e. the context IRI) of such links to be the IRI that identifies the representation which is provided as the content of the corresponding 2xx HTTP response. The rationale for this proposal is the following. According to the Linked Data principles I may obtain data about a thing by looking up the URI for that thing; what I get is a set of RDF triples extracted from a representation of a resource about the thing. In order to assess the quality and trustworthiness of these triples I don't (only) want general provenance information about the resource; instead, I'm particularly interested in the provenance of the specific representation from which I extracted the triples.
BTW, my proposal does not rule out linking to a provenance description for the resource. It is possible to explicitly use the anchor parameter to explicitly specify the context IRI of a link with rel="provenance".
Olaf
[1] http://tools.ietf.org/html/rfc5988#section-5.2
- Related Actions Items:
- No related actions
- Related emails:
- Reviewing outstanding issues on PAQ document. (from GK@ninebynine.org on 2011-09-15)
- Provenance context URIs for RDF data (was: PAQ document update, target renamed as context) (from GK@ninebynine.org on 2011-08-25)
- Re: PAQ document update, target renamed as context (from GK@ninebynine.org on 2011-08-25)
- Re: PAQ document update, target renamed as context (from GK@ninebynine.org on 2011-08-22)
- Re: PAQ document update, target renamed as context (from hartig@informatik.hu-berlin.de on 2011-08-21)
- Re: PAQ document update, target renamed as context (from GK@ninebynine.org on 2011-08-19)
- Re: PROV-ISSUE-68 (http-link-domain): Domain of HTTP links with rel=provenance [Accessing and Querying Provenance] (from GK@ninebynine.org on 2011-08-04)
- PROV-ISSUE-68 (http-link-domain): Domain of HTTP links with rel=provenance [Accessing and Querying Provenance] (from sysbot+tracker@w3.org on 2011-08-02)
Related notes:
When using the HTTP Link: header, this might be specified explicitly by the "anchor" parameter.
When using the HTML <link> element, there is no equivalent mechanism. A new link relation type has been introduced which addresses the problem of finding *a* context-URI for the provenance, but that is not the same as giving an explicit subject/domain for the link itself.
My tentative response to this question is: the domain is the containing or retrieved resource. So for resources retrieved using HTTP, that is the URI of the retrieved resource. For other resources, the URI might be unknown, corresponding to a blank node in RDF.
All this is suggesting to me that the "anchor" link relation be renamed "context", clearing the way for the HTTP Link: anchor= param to be used as defined, without necessarily affecting the interpretation of provenance context.
The down-side of this is that it introduces a subtle and possibly confusing distinction between the "context" of an HTTP Link: header and the provenance context associated with a resource.Graham Klyne, 19 Aug 2011, 16:41:25
The previous comment also concerns a comment in the email http://lists.w3.org/Archives/Public/public-prov-wg/2011Aug/0237.html concerning section 3.1
Graham Klyne, 19 Aug 2011, 16:44:05I think this is now covered by updated text in 3.1 and elsewhere, and the clarified use of "entity-URI".
Graham Klyne, 1 Dec 2011, 13:21:55Closing
Graham Klyne, 5 Jan 2012, 11:11:37Display change log
