- From: Karen Coyle <kcoyle@kcoyle.net>
- Date: Thu, 1 Mar 2018 07:50:34 -0800
- To: public-dxwg-wg@w3.org
Rob, we can put this on next week's plenary agenda if you feel ready to discuss and vote on it. It will need one or more stated requirements; those could come out of the plenary discussion if you think that is best. kc On 3/1/18 4:19 AM, Rob Atkinson wrote: > > Good question - i am thinking mainly about the case where the container > and the contents need to be described (a list) and making sure we can > handle this - and also the need to tell people that a link is safe to > follow (an item) - or a commitment to download an entire data set and > find the item. This initial distinction has quite of lot of extra > utility in that regard and can handle quite general cases in the spirit > of dcat. > > i..e certainly not as specific as GraphQL - i think to go to a finer > grained distinction we would need another use case or more evidence of > need and practice. > > So if the library desk has a URI identifier and can deliver a result set > represented with an identified profile of some spec, and packaged in an > identifiable container - then why not? > > Rob > > > > On Thu, 1 Mar 2018 at 22:43 pedro winstley > <pedro.win.stan@googlemail.com <mailto:pedro.win.stan@googlemail.com>> > wrote: > > Hi Rob > > In your item #2 about the nature of the deliverable/s from the > 'endpoint', what assumptions are being made about the nature (e.g. > transport protocol) for this 'endpoint'. I mean, could it be a > library desk, for example; or an xmlrpc call? Or are we just > imagining HTTP endpoints with REST/HATEOAS , or perhaps GraphQL > endpoints? > > Peter > > On 28 February 2018 at 22:10, Rob Atkinson <rob@metalinkage.com.au > <mailto:rob@metalinkage.com.au>> wrote: > > When considering Linked Data applications of dataset > descriptions, and content negotiation by profile, there is a > need to distinguish between distributions that package the > entire dataset - and hence may have a "container" structure, and > service endpoints that deliver a specific object. > > So if we had a metadata catalog that supported a query service > that returned a set of DCAT records (using say the BotDCAT-AP > profile), and also an api that delivered a specific record using > this: > > 1) we would need to be able to specify the query service support > the BotDCAT-AP profile and MyDCATCatalogSearchFunction profile > (that specifies how the container structure is implemented?) > > > 2) we need to distinguish between enpoints that > a) deliver a single item > b) deliver a list based on some filter/criteria > c) deliver the entire dataset > > case c) is covered by dcat:downloadURL but dcat:accessURL is > too general to distinguish between the others. we need one of: > i.) canonical subclasses of accessURL or item and list endpoints > ii.) a type property (dct:type?) for the distribution with > canonical values for list and endpoint cases > iii) a property e.g. dcat:containerProfile to indicate that > data conforming to a domain profile will be wrapped in a > container conforming to an access method profile. > > Note - I will be able to provide an implementation of whatever > method is chosen - currently having to create my own vocab to > annotate links these ways - so trivially easy to update to DCAT > support for these notions. > > Should I also formally submit this as a comment as feedback to > the UCR - or is this sufficient to process it as such. > > Apropos of this, I have a little profiles vocabulary in > preparation I will post during the week that will allow short > tokens to be association with profiles for use in access > endpoints and profile negotiation, whether this vocabulary is a > standandalone and subclasses dcat:accessURL or whether all or > part could be rolled into dcat is an open question. It would > provide a canonical means for dcat to specify profile > conformance, but adds a little structure. > > Maybe using the modular approach discussed I should join the > DCAT sub group and propose it, with a fall back of a separate > namespace? > > Rob Atkinson > > -- Karen Coyle kcoyle@kcoyle.net http://kcoyle.net m: 1-510-435-8234 (Signal) skype: kcoylenet/+1-510-984-3600
Received on Thursday, 1 March 2018 15:51:12 UTC