We must consider all use cases. (1) the KISS case you present is the easy one: URIs natively map to their URLs. (2) the redirection case, with admin rights:: Oops, i had to rename my files on the server and now my URIs no longer match their URLs. Fortunately, I have access to a redirection feature (symbolic links on the server, .htaccess for 303, mod_rewrite, etc). (3) the redirection case, without admin rights: Idem as (2) but no workaround available. The URIs-URLs scheme is *definitely* broken. And i am now a happy provider of 404 errors.
We need to establish *one* best practice to manage all 3 cases, with only minor additional work when you need to switch from one case to another. Discarding cases (2) and (3), or considering that 404 is a non-issue are, imho, very short-sighted positions. But I strongly agree with you that "simple things should be simple". Let's not forget the second part of the mantra: "but complex things should be possible". That is why we need to find the *one* recipe for all 3 cases. On Friday, July 10, 2009, Hugh Glaser <[email protected]> wrote: > Thank you all for not (yet) incinerating me. > Some responses: > > I'm not really fussed about html documents - to me they aren't really "in" > the semantic web, other than the URL is a string, which can be interpreted as > something that can use the same access mechanisms as my URIs to find some > more text. I do publish html documents of my RDF, but that is only to permit > aficionadas to browse the raw data. > If I actually have html documents, then something like RDFa is probably a > great way of doing things. > > Many people worry about the modelling, which is great and why RDF is so good. > But I start more from the consumer's end rather than the modeller's, and work > back through the publisher. > Does anyone actually have a real application (and I am afraid I don't really > count semantic web browsers as applications) that has a problem getting the > RDF if I have a file at > http://wrpmo.com/hg > which contains > <http://wrpmo.com/hg> <http://xmlns.com/foaf/0.1/name> "Hugh Glaser" . > <http://wrpmo.com/hg> <http://www.aktors.org/ontology/portal#has-web-address> > "http://www.ecs.soton.ac.uk/people/hg"; . > and I then use http://wrpmo.com/hg as one of "my" URIs? > Certainly doesn't bother my applications. > And your average enthusiastic sysprog or geek can understand and do this, I > think - that's why RDFa is getting popular. > I know that things like dc:creator can be a little problematic, but we are > paying a high price, I think. > > Steve's comments on using vi are interesting. > Yes, we used vi and hackery. > In fact I still generate those old web pages by running a Makefile which > finds a .c source and calls the C preprocessor to generate the .html pages, > and I certainly started this in the early 90s. > At the moment I use all sorts of hackery to generate the millions of triples, > but the deployment is complex. > > Is it really such a Bad Thing if I do http://wrpmo.com/hg, if the alternative > is that I won't publish anything? > Surely something is better than nothing? > In any case, just like html browsers, linked data consumers should deal with > broken RDF and get the best they can out of it, as going back and telling the > server that the document was malformed, or reporting to a "user" is no more > an option in the linked data world than it is in the current web. > > Of course, as a good citizen (subject? - footsoldier?) of linked data and the > semantic web, I hope I do all the stuff expected of me, but it doesn't mean I > think it is the right way. > > Thank you very much for the considered responses to such an old issue. > Best > Hugh > > On 10/07/2009 11:13, "Steve Harris" <[email protected]> wrote: > > On 10 Jul 2009, at 10:56, Richard Light wrote: >> In message <[email protected]>, Steve >> Harris <[email protected]> writes >>> On 10 Jul 2009, at 01:22, Hugh Glaser wrote: >>>> If I can't simply publish some RDF about something like my dog, by >>>> publishing a file of triples that say what I want at my standard >>>> web site, >>>> we have broken the system. >>> >>> I couldn't agree more. >>> >>> <rant subject="off-topic syntax rant of the decade"> >>> Personally I think that RDF/XML doesn't help, it's too hard to >>> write by hand. None of the other syntaxes for RDF triples really >>> have the stamp of legitimacy. I think that's something that could >>> really help adoption, the the same way that strict XHTML, in the >>> early 1990's wouldn't have been so popular with people (like me) >>> who just wanted to bash out some text in vi. >>> </> >> >> Well, in my view, when we get to "bashing out" triples it isn't the >> holding syntax which will be the main challenge, it's the Linked >> Data URLs. Obviously, in a Linked Data resource about your dog, you >> can invent the URL for the subject of your triples, but if your Data >> is to be Linked in any meaningful way, you also need URLs for their >> predicates and objects. >> >> This implies that, without a sort of Semantic FrontPage (TM) with >> powerful and user-friendly lookup facilities, no-one is going to >> bash out usable Linked Data. Certainly not with vi. And if you >> have such authoring software, the easiest part of its job will be >> rendering your statements into as many syntaxes as you want. > > I think that's a fallacy. I the web wasn't bootstrapped by people > wielding Frontpage*. It was people like Hugh and I, churning out HTML > by hand (or shell script often), mostly by "cargo cult" copying > existing HTML we found on the Web. That neatly sidesteps the schema > question, as people will just use whatever other people use, warts, > typos, and all. > > The tools for non-geeks phase comes along much later, IMHO. First we > have to make an environment interesting enough for non-geeks to want > to play in. > > Happy to be demonstrated wrong of course. > > - Steve > > * Frontpage wasn't released until late '95, and wasn't widely known > until late '96 when it was bought by MS. By which time the Web was a > done deal. > > -- > Steve Harris > Garlik Limited, 2 Sheen Road, Richmond, TW9 1AE, UK > +44(0)20 8973 2465 http://www.garlik.com/ > Registered in England and Wales 535 7233 VAT # 849 0517 11 > Registered office: Thames House, Portsmouth Road, Esher, Surrey, KT10 > 9AD > > > > >
