- From: Noah Mendelsohn <nrm@arcanedomain.com>
- Date: Fri, 14 Dec 2012 17:46:18 -0500
- To: Robin Berjon <robin@w3.org>
- CC: "Eric J. Bowman" <eric@bisonsystems.net>, "Martin J. Dürst" <duerst@it.aoyama.ac.jp>, Karl Dubost <karld@opera.com>, Paul Libbrecht <paul@hoplahup.net>, "www-tag@w3.org List" <www-tag@w3.org>
On 12/14/2012 11:37 AM, Robin Berjon wrote: > > No, a media type is just as bad, only one step removed. If the > implementation is in HTML (as it's likely and getting likelier to be) then > all you've done is pushed the masquerading to another layer. I'm not convinced. Media types and schemes affect different parts of the space. Schemes show up directly in the names that are used for linking, and they require hooks in the corresponding dispatch tables. When you use a scheme other than http, you cannot change your mind about the formats or protocols to be used without changing all the URIs and links. With http, you can have a link to a portion of a map (to use the example we're talking about), and have it come back as an interactive application sometimes, and a gif others. My impression is that Google Maps does exactly this in supporting small devices with non-JavaScript browsers. I think the architectural differences between scheme and media type are significant, and like most important architectural principles, they translate into use cases that are important. AWWW says some about this in section 2.4 [1], and also references rfc2718 [2]. Section 4.1 of the TAG Finding "Self Describing Web" [3] makes the case a little more strongly, pointing out that URIs and Protocols are "deeper" in the stack than media types; usually more software has to be changed in more places to get universal support for a new schema than for a media type. Noah [1] http://www.w3.org/TR/webarch/#URI-scheme [2] http://www.ietf.org/rfc/rfc2718.txt [3] http://www.w3.org/2001/tag/doc/selfDescribingDocuments.html#stablelayers
Received on Friday, 14 December 2012 22:47:03 UTC