- From: Alexandre Bertails <bertails@w3.org>
- Date: Thu, 08 Nov 2012 08:29:23 -0500
- To: Roger Menday <roger.menday@uk.fujitsu.com>
- CC: Henry Story <henry.story@bblfish.net>, "Wilde, Erik" <Erik.Wilde@emc.com>, "public-ldp-wg@w3.org" <public-ldp-wg@w3.org>, nathan <nathan@webr3.org>
On 11/08/2012 05:40 AM, Roger Menday wrote: > > On 8 Nov 2012, at 08:25, Henry Story wrote: > >> >> On 8 Nov 2012, at 00:56, "Wilde, Erik" <Erik.Wilde@emc.com> wrote: >> >>> hello henry. >>> >>> On 2012-11-07 15:27 , "Henry Story" <henry.story@bblfish.net> wrote: >>>> On 8 Nov 2012, at 00:12, "Wilde, Erik" <Erik.Wilde@emc.com> wrote: >>>>> that's what on the web media types are doing. i know that this is way >>>>> outside of the scope of this group, but since we're saying REST in the >>>>> charter, this is what we would be doing in a RESTful design: design a >>>>> media type that represented the concepts we're building interactions >>>>> around, and then making the distinction you're pointing out is done by >>>>> virtue of the media type. >>>> I think you are trying to put too much in the media types. The Media type >>>> is just a way to interpret a document - i.e. to extract its semantics. >>> >>> nope, it's more than that. it defines the set of interconnected resources >>> a client can traverse, and defines that traversing this set of resources >>> means. for every link that a client can find, the media type specifies why >>> a client might want to follow that link, and maybe what a client has to do >>> when following that link. >> >> You can do that with RDF too, you just choose special vocabularies instead >> of choosing special mime types. > > I agree with that. we don't want to go the way of many REST apis where a new mediatype is defined for each (XML schema) type in the system. I think that one mime type will be enough for LDP. The cost of looking at the RDF to decide what to do may be pretty high. Alexandre. > > Roger > >> >> >>> >>>> >>>>> yup, and that would be the header signaling the media type. >>>> As said above that would be like saying that servers MUST speak a >>>> different >>>> language from the other documents they are serving, which seems arbitrary. >>> >>> it's the opposite. it's the difference in functionality that's exposed as >>> media types. >> >> That's a mistake, that just happens to work. >> >>> if you are an XML database, you accept any XML and just store >>> it. that's fine. if you also allow people to interact with any kind of >>> management functionality of the database, what you exchange is still XML, >>> but its meaningful (let's say some XACML for managing access right) and >>> thus labeled by a media type that makes that distinction clear. that's >>> just how HTTP works. >> >> Http allows you to do content negotiation on a resource to get back >> a preferred representation of that resource. All representations returned >> should be pretty much equal. That is where the idea of semantics comes from: >> there is something all these representations have in common. >> >> What you are describing is in my view just a lucky error that people on >> REST mailing lists have used because it seems enough like it solves the >> problem, when in fact it just makes things more complicated. For example >> that way of working makes things a lot more complicated as all of a sudden >> you have to create a whole syntax for servers to work with, just to >> distinguish when the server is speaking from when the document is served by >> it but is not a statement made by the server. >> >> That solution is at the wrong place at the logical layer. What you want is >> information about WHO said something, and the solution you are describing >> is telling me HOW it is said. Then there is a backchannel convention of which >> actors can say something which way to get to the WHO. >> >> Much simpler would be to at least start out by thinking about WHO is >> saying something, since the original problem was at that layer. Is the >> server telling me that this is a collection? Or is this just a document >> someone else wrote saying it is a collection? >> >> In any case on could also just argue: don't put a document saying >> >> <> a ldp:Container >> >> anywhere. It would be like putting up a web page that was lying, and people >> will end up removing links to that resource, and distrusting servers that >> publish it. If one wanted to help servers publish documents of people on the >> web they did not fully control, then it would be useful to allow the server to >> say that it is not responsible for what is in the document. >> >>> >>> cheers, >>> >>> dret. >>> >> >> Social Web Architect >> http://bblfish.net/ >> >
Received on Thursday, 8 November 2012 13:29:49 UTC