- From: Corey Harper <corey.harper@gmail.com>
- Date: Fri, 25 Jan 2013 22:07:41 -0500
- To: "Young,Jeff (OR)" <jyoung@oclc.org>
- Cc: Kevin Ford <kefo@3windmills.com>, public-schemabibex@w3.org
Hi Jeff, Definitely not. My apologies if I gave that impression. I completely agree that, when it fits the spirit of a well-constructed model, resources are preferable. That said, I have a hard time with a proliferation of standards that *require* a range of resource when doing so presents a barrier to people's pragmatic efforts to publish data. I've made this same argument with regard to dcterms: for years now. I just don't see the downside to Kevin's suggestion that we might "permit a string OR rdf:Resource as the range of the inStandard/codingScheme property, and we encourage the use of the latter". Thanks, -Corey On Fri, Jan 25, 2013 at 9:47 PM, Young,Jeff (OR) <jyoung@oclc.org> wrote: > Corey, > > As Kevin pointed out, Karen misunderstood me. Are you suggesting that strings really are better than things and that Schema.org would agree? > > http://googleblog.blogspot.com/2012/05/introducing-knowledge-graph-things-not.html > > As discussed in another thread, I have no problem with developers ignoring httpRange-14 and Cool URIs or the use cases that justify them. If the Web was replaced tomorrow I wouldn't fret because the models being developed transcend the Web. > > Jeff > >> -----Original Message----- >> From: Corey Harper [mailto:corey.harper@gmail.com] >> Sent: Friday, January 25, 2013 8:53 PM >> To: Young,Jeff (OR) >> Cc: Kevin Ford; public-schemabibex@w3.org >> Subject: Re: question about medical code >> >> Hi Jeff, et al., >> >> I'm inclined to agree with Karen here. I think a lot of libraries are >> more trying to solve practical, near-term problems with schema.org and >> may not be in a position to adopt the kinds of "proper linked data" >> practices that we'd want in an ideal world. Setting aside your "if they >> have access to the HTML production", which is not usually the case with >> the current generation of ILS systems, even if folks did, we have a >> tendency to set a pretty high barrier for that. Many of the libraries I >> talk to are frankly scared of the prospect of minting "Cool URIs". >> There's so many rules, and the semweb community can be rather pedantic >> about them. >> >> People are scared of the notion that Cool URIs never change (really, a >> redirect isn't a fair solution to that problem?) They're set in stone >> for how long? The rhetoric makes people feel like they have to commit >> to a URI for quite longer than we've even had a Web. Plus, there's all >> the details of hash URIs vs. slash URIs with proper redirects between >> things & pages that describe them. And I could go on. And on. >> >> But the point is that each of these requirements serves as a barrier to >> libraries making their data available in a more machine usable form >> than it exists now. Do we really want to say that if you're not in a >> position to create clean, cruft-free, properly-range-14-compliant URIs >> for everything you might wish to describe than we actually don't want >> your structured data at all? >> >> What I see as the promise of schema.org is it's potential to *lower* >> the bar for publishing the data folks have, using the systems they have >> available. >> >> Sorry for the rant, but I worry that we're going to (again) shoot >> ourselves in the foot by putting up roadblocks that very few in our >> profession (and very few web developers *outside* of libraries) have >> the time or energy to surmount. >> >> Thanks, >> -Corey >> >> On Fri, Jan 25, 2013 at 8:03 PM, Young,Jeff (OR) <jyoung@oclc.org> >> wrote: >> > Kevin's got it. I would like to imagine schema:ConceptScheme (or >> > something similar) instead of rdfs:Resource as the range for this >> > property, but that's merely a detail. >> > >> > Coining proper Linked Data (hash) URIs matched up to some RDFa should >> > be very easy if they have access to the HTML production. >> > >> > Once again, remember that Schema.org is automatically forgiving when >> > people put strings where things are expected. That's true across the >> > board. It shouldn't be necessary to think this property is any >> > different. >> > >> > Jeff >> > >> >> -----Original Message----- >> >> From: Kevin Ford [mailto:kefo@3windmills.com] >> >> Sent: Friday, January 25, 2013 7:29 PM >> >> To: public-schemabibex@w3.org >> >> Subject: Re: question about medical code >> >> >> >> Dear Karen, >> >> >> >> Jeff was referring to the codingSystem property at [1], not the code >> >> itself. (Richard proposed an alternate name for the property - >> >> inStandard - in an above example.) >> >> >> >> Both examples, however, represent the scheme to which the identifier >> >> conforms as simple strings. Jeff, I believe, would prefer if the >> >> scheme to which the identifier conforms were referenced by URI. As >> > in: >> >> >> >> <http://bowker.com/identifiers/isbn/9780553479430> >> >> a schema:Identifier; >> >> schema:name "9780553479430"; >> >> schema:inStandard >> >> <http://id.loc.gov/vocabulary/identifiers/isbn> >> >> . >> >> >> >> Now, it might be that a string value *or* URI would be perfectly >> >> acceptable for the property schema:inStandard, but what I really >> > wanted >> >> to comment on is slightly tangential because it involves the choices >> > in >> >> modelling before you/us. >> >> >> >> schema:inStandard rdfs:range xsd:string >> >> >> >> is a popular and easy thing to do because it allows anyone to plug >> in >> >> their value. However >> >> >> >> schema:inStandard rdfs:range rdf:Resource >> >> >> >> can lead to more information and better control. >> >> >> >> The former is easy for anyone to understand and produce, because >> they >> >> simply need to enter some text. The latter, however, would require >> >> someone with a "special" identifier (i.e. an identifier type that >> has >> >> not been given its own URI) needing to mint a URI. >> >> >> >> That's not that hard for us to do or necessarily understand, but I >> > find >> >> that this is one of those recurring issues with respect to Linked >> > Data. >> >> People have something custom, there's no URI, and he doesn't know >> >> what to do next. >> >> >> >> I don't have a real solution to offer (unless we permit a string OR >> >> rdf:Resource as the range of the inStandard/codingScheme property, >> >> and we encourage the use of the latter) but I didn't want to miss >> the >> >> opportunity to note that this is one of those issues that will recur >> >> and, however it pans out, requires some form of education. >> >> >> >> Yours, >> >> Kevin >> >> >> >> [1] >> >> >> > >> http://www.w3.org/community/schemabibex/wiki/Identifier#Adopting_appro >> > a >> >> ch_from_medical.2Fhealth_extension >> >> >> >> >> >> >> >> On 01/25/2013 07:04 PM, Karen Coyle wrote: >> >> > Jeff, I believe that the issue is that the codes *are* strings, >> and >> >> > that's why they need this two-part description. If they were >> things >> >> we >> >> > wouldn't be having this conversation. >> >> > >> >> > kc >> >> > >> >> > On 1/25/13 12:29 PM, Young,Jeff (OR) wrote: >> >> >> I think they should model values for codingScheme as things >> rather >> >> than >> >> >> as strings. That would be the SKOS way. They already have a >> >> forgiveness >> >> >> clause for numbing things down to strings, but it's harder to row >> > in >> >> the >> >> >> opposite direction. >> >> >> >> >> >> Jeff >> >> >> >> >> >>> -----Original Message----- >> >> >>> From: Karen Coyle [mailto:kcoyle@kcoyle.net] >> >> >>> Sent: Friday, January 25, 2013 3:13 PM >> >> >>> To: public-schemabibex@w3.org >> >> >>> Subject: question about medical code >> >> >>> >> >> >>> Adrian has added the medical code to the Identifier page [1]. It >> >> looks >> >> >>> to me like in its simplest form it could also be used for the >> >> >>> minimalist approach to identifiers that I have proposed [2]. >> >> >>> Essentially, it only needs two properties: >> >> >>> >> >> >>> codeValue Text The actual code. >> >> >>> codingSystem Text The coding system, e.g. 'ICD-10'. >> >> >>> >> >> >>> Of course, they have to be grouped as a single unit, which the >> >> medical >> >> >>> code page calls "code": >> >> >>> >> >> >>> code MedicalCode A medical code for the entity, taken >> from >> >> a >> >> >>> controlled vocabulary or ontology such as ICD-9, DiseasesDB, >> >> >>> MeSH, SNOMED-CT, RxNorm, etc. >> >> >>> >> >> >>> I must admit that I find the properties here to be a bit >> circular >> >> but >> >> >>> I'm going to assume that greater minds than mind have >> >> >>> investigated >> >> >> this >> >> >>> and determined that it works. >> >> >>> >> >> >>> I could add an example on my simplified identifier page, and/or >> >> could >> >> >>> add a simplified example after Adrian's. Does that make sense? >> >> >>> >> >> >>> I have one worry about using "code" however: I think that we, >> >> >>> too, >> >> >> will >> >> >>> have codes that need to be described in this way. Will there >> need >> >> to >> >> >> be >> >> >>> a difference between codes of this type and identifiers? It >> seems >> >> to >> >> >> me >> >> >>> that folks are often using "identifier" to me an identifier for >> > the >> >> >>> focus of the description, whereas "code" could be, for example, >> a >> >> >>> description element like "audience level" or "government >> document >> >> >>> type." >> >> >>> In practical usage, will we need both? >> >> >>> >> >> >>> kc >> >> >>> -- >> >> >>> Karen Coyle >> >> >>> kcoyle@kcoyle.net http://kcoyle.net >> >> >>> ph: 1-510-540-7596 >> >> >>> m: 1-510-435-8234 >> >> >>> skype: kcoylenet >> >> >>> >> >> >> >> >> >> >> >> >> >> >> >> >> >> > >> >> >> > >> > >> > >
Received on Saturday, 26 January 2013 03:08:13 UTC