- From: Wilde, Erik <Erik.Wilde@emc.com>
- Date: Thu, 28 Feb 2013 11:02:12 -0500
- To: "public-ldp-wg@w3.org" <public-ldp-wg@w3.org>
- CC: John Arwe <johnarwe@us.ibm.com>
hello john. On 2013-02-25 21:54 , "John Arwe" <johnarwe@us.ibm.com> wrote: >If there is a more appropriate existing word that already fits perfectly >an existing consensus within the WG, sure fine. I'd be lying if I said I >believed any such word already existed - establishing shared meanings is >inherently > a social process that requires shared experiences, and I have yet to see >evidence of widespread durable pre-existing consensus within the members >of this WG. Often a strawman (and some matches) are a quick way to get >there however ;-) i like that approach. let's look at a simple example: affordances in HTML. HTML has links that matter at various levels, and it makes sense for the media type to clearly talk about all of them. some are "within the media type" and thus allow user agents to do smart things automatically. others are "leaving the media type" and thus make more sense as a something that a user agent makes available to the user to decide. - within media type: <img/> by definition links to an image of some sort, so the media type still matters in the sense that HTML says that "an image should be displayed in a visual user agent." so this is something that a user agent can do by itself. - leaving the media type: <a> can lead anywhere, including pretty much random URI schemes, so all a user agent can do is present that affordance to the user. so this matter as well, of course, but is at a different level. now, for collection/entries, we probably have the same distinction. being able to find from a member to a collection or the other way matters, because then we can build code that crawls collections, or that starts from entries and follows links from there. also, by using a self-contained model for entries, we can build processes where clients find entries, then find collections, then crawl collections and stuff the entries into new collections they are assembling from random assortments of entries. all very useful and only possible if clients can use affordances to engage in all these interactions. and then entries might have all sorts of other links, which we may still want to expose, but we're not constraining what they link to. in these cases, user agents can still expose these links ("click here to find ..."), but it would be outside of the scope of LDP to say what is expected as a URI scheme, or as a media type. cheers, dret.
Received on Thursday, 28 February 2013 16:03:21 UTC