Closed VladimirAlexiev closed 9 years ago
It is explicitly stated that void:sparqlEndpoint
must not be used:
We recommend that the void:sparqlEndpoint MUST NOT be used since it would either generate a maintenance issue for the generated descriptions be ambiguous as to which version of the data is loaded in the triplestore.
What maintenance issues? That you'd have to update a statement for every version? What's the big problem with this? And even if you leave the statement obsolete, that's still better than not telling the user where to find the SPARQL endpoint (a vital piece of information).
(BTW the above sentence doesn't parse properly)
The sentence is grammatically wrong indeed. That is something I can fix.
For your in-depth question, please get in touch with the other contributors/editors on the HCLS mailing list (public-semweb-lifesci@w3.org) or join our conference calls (11AM ET / 4PM UK / 5PM CET, +1 617 761 6200, irc.w3.org port 6665 channel "#HCLS").
Thank you for creating these issues, because it helps us improve our draft.
@VladimirAlexiev Please revisit by previously quoted paragraph. I have split the sentence into two sentences. I hope our intention is clearer now (or, clear for the first time).
Hi Joe, that's fine, but whatever reasons you have for witholding the void:sparqlEndpoint, are wrong reasons. (Sorry, I can't engage in discussion on that mlist.)
The reason we don't include a sparqlEndpoint from the dataset description is because that version of the dataset may not actually be in the endpoint. i think a better way is that the endpoint describes what data it contains, as per http://www.w3.org/TR/sparql11-service-description/
If the data is not in an endpoint, section 6.5.3 does not apply at all.
Sorry Joe, I don't mean to troll. I've spent some time making a comprehensive description of the Getty vocabs (http://vocab.getty.edu/doc/#Descriptive_Information), so I think I know a bit.
Information about endpoints and their contents are the subject of a different guideline.
the recommendation of dct:source is a product of consensus in this effort.
On Wed, Sep 3, 2014 at 8:48 AM, Vladimir Alexiev notifications@github.com wrote:
How would the user find that endpoint to get that SD that points to the dataset version?
And even if they found it, your use of dct:source is not recommended in http://www.w3.org/TR/void/#sparql-sd or any other specification..
— Reply to this email directly or view it on GitHub https://github.com/joejimbo/HCLSDatasetDescriptions/issues/80#issuecomment-54317728 .
Is an alternative to this to put void:sparqlEndpoint on the summary level description so that it is not tied to a version?
On Wed, Sep 3, 2014 at 6:19 PM, Michel Dumontier notifications@github.com wrote:
Information about endpoints and their contents are the subject of a different guideline.
the recommendation of dct:source is a product of consensus in this effort.
On Wed, Sep 3, 2014 at 8:48 AM, Vladimir Alexiev notifications@github.com
wrote:
How would the user find that endpoint to get that SD that points to the dataset version?
And even if they found it, your use of dct:source is not recommended in http://www.w3.org/TR/void/#sparql-sd or any other specification..
— Reply to this email directly or view it on GitHub < https://github.com/joejimbo/HCLSDatasetDescriptions/issues/80#issuecomment-54317728>
.
— Reply to this email directly or view it on GitHub https://github.com/joejimbo/HCLSDatasetDescriptions/issues/80#issuecomment-54322607 .
M. Scott Marshall, PhD MAASTRO clinic, http://www.maastro.nl/en/1/77/strategy-plan.aspx http://radiomics.org http://eurecaproject.eu/ http://semantic-dicom.org/ http://www.linkedin.com/pub/m-scott-marshall/5/464/a22
i'm fine with that.
On Mon, Sep 15, 2014 at 8:22 AM, mscottm notifications@github.com wrote:
Is an alternative to this to put void:sparqlEndpoint on the summary level description so that it is not tied to a version?
On Wed, Sep 3, 2014 at 6:19 PM, Michel Dumontier notifications@github.com
wrote:
Information about endpoints and their contents are the subject of a different guideline.
the recommendation of dct:source is a product of consensus in this effort.
On Wed, Sep 3, 2014 at 8:48 AM, Vladimir Alexiev < notifications@github.com>
wrote:
How would the user find that endpoint to get that SD that points to the dataset version?
And even if they found it, your use of dct:source is not recommended in http://www.w3.org/TR/void/#sparql-sd or any other specification..
— Reply to this email directly or view it on GitHub <
https://github.com/joejimbo/HCLSDatasetDescriptions/issues/80#issuecomment-54317728>
.
— Reply to this email directly or view it on GitHub < https://github.com/joejimbo/HCLSDatasetDescriptions/issues/80#issuecomment-54322607>
.
M. Scott Marshall, PhD MAASTRO clinic, http://www.maastro.nl/en/1/77/strategy-plan.aspx http://radiomics.org http://eurecaproject.eu/ http://semantic-dicom.org/ http://www.linkedin.com/pub/m-scott-marshall/5/464/a22
— Reply to this email directly or view it on GitHub https://github.com/joejimbo/HCLSDatasetDescriptions/issues/80#issuecomment-55606355 .
So the recommendation is
Property | Value | Summary | Version | Distribution |
---|---|---|---|---|
void:sparqlEndpoint | IRI | SHOULD | NEVER | NEVER |
And in the guidance text we need to explain that the endpoint should be consulted to discover which distribuion file of which version of a dataset is currently loaded, i.e. the example that we already have.
Decision reached during the call 22 September 2014
Property | Value | Summary | Version | Distribution |
---|---|---|---|---|
void:sparqlEndpoint | IRI | SHOULD | SHOULD NOT | SHOULD NOT |
Strengthen the guidance text to explain why it is problematic, e.g. you would need to ensure that you persist this SPARQL endpoint with this data for the future.
done. 3d3cd37c41dfd185fdfe23fdec733533707a8b9e
Surely this section should use void:sparqlEndpoint to point to the endpoint. (I've also used dcat:accessURL to point from an RDF distribution to its endpoint, but not sure that's useful)