- From: Thomas Baker <tbaker@tbaker.de>
- Date: Mon, 6 Feb 2006 08:31:43 +0100
- To: SW Best Practices <public-swbp-wg@w3.org>
Telecon: SWBPD VM TF Note: was scheduled for 2006-01-31 but actually held 2006-02-01. IRC: http://www.w3.org/2006/02/01-vmtf-irc Attendees: Ralph Swick, Alistair Miles, Tom Baker (chair) Agenda: http://lists.w3.org/Archives/Public/public-swbp-wg/2006Jan/0180.html Previous: 2006-01-24 http://lists.w3.org/Archives/Public/public-swbp-wg/2006Jan/0125.html Topic: Document Title RESOLVED: Best Practice Recipes for Publishing RDF vocabularies Topic: removing content types RalphS and Tom concur to removing text/xml and application/xml as conditions from the rewrite rules. RESOLVED: to remove them. Alistair notes: http://www.w3.org/TR/owl-ref/#MIMEType mime type for OWL -- MIME Type section in OWL Reference Guide: 'application/xml' is allowed as mime type for owl Ralph: we should propose Best Practice, not "all acceptable practices". Discussion follows... Alistair: SKOS and DC - stuff is in quite different places Ralph: If they are on different servers, cannot use a type map. Type map only works if everything in directory has the same quality rules, applies to everything in directory tree. Is there a way to configure the whole host? If always prefer to serve HTML over x... without having to do a type map for each URI...? Al: two use-cases here. Simple one: when server doesn't specify any difference in quality of representations, but client does: "I prefer HTML to RDF but I can accept both". In that case, cond'l redirects in order. Additional situation: server wants to specify some diff in quality... But why would a server want to do that? Ralph: in my case - maybe not "real-life" - what happens if client does not include any accept header - has no preference. The specific rules take care of this - falls through to the end. When I tried this, I wound up not using conditional rewrites -- why did this feel like the best solution? Maybe write a rewrite condition that says "if there is no accept header"...? I.e., doing a conditional rewrite for a missing Accept...? We should think more about case where client can accept both. But not a pressing question... Not important enough to delay publication. New issue came up: http://lists.w3.org/Archives/Public/public-swbp-wg/2006Jan/0185.html re: RDF/A Primer Version. Ben Adida summarized yesterday: If you don't have a hash in your URI, then you cannot then resolve to URIs with hashes that do not correspond to a fragment of a document. Is this a problem? Alistair: recipes specifically dodge this. Nothing is ever served directly from the namespace URI - always a redirect from somewhere else. Therefore, no Web interaction will ever tell you that the namespace URI denotes a document. I believe if people use these recipes they put themselves in the safest possible corner w.r.t. this issue of information resources vs. non-information resource. Example: http://purl.org/dc/elements/1.1/title resolves to http://dublincore.org/2003/03/24/dces#title (the RDF schema). The HTML expression is at http://dublincore.org/documents/dcmi-terms/#title. (For DC, both text/html and application/rdf+xml _should_ be 303 redirects...) So DC would not have a problem here. Alistair has the edit token to remove the extra mime type redirects, then will pass the token back to Tom to send mail to the WG. Discussion of @@'s in editor's draft -- @@TODO verify minimal required overrides Alistair: I removed some overrides that Ralph added, verified these work, but haven't done exhaustive testing -- see recipe 6 (@@TODO). Alistair: I plan to leave both of these -- @@TODO why. in Content Negotiation ACTION: Ralph write text on why examining user-agent is a hack. (Doesn't need to stop publication, however.) -- [[ Accept: application/rdf+xml, ??? @@TODO ]] in same section Alistair: mail from Jeremy on this. We should decide whether application/xml should be in this list. There's no mime type for Turtle, N3, etc. Ralph: I propose we stick with just mime types for W3C Recommendations so this would leave us with just application/rdf+xml and text/html. Today, if they're only interested in RDF, they wouldn't need text/html. In the future, if GRDDL becomes a Rec, then a large class of HTML documents will become carriers for RDF. Alistair: propose Accept: application/rdf+xml, application/xml. Ralph: yes, because OWL endorses application/xml too. Alistair: how about , */* ? Ralph: no, we're giving advice specifically to "applications that expect to process RDF content". Eventually, when we have GRDDL, we can change this. -- Dr. Thomas Baker baker@sub.uni-goettingen.de SUB - Goettingen State +49-551-39-3883 and University Library +49-30-8109-9027 Papendiek 14, 37073 Göttingen
Received on Monday, 6 February 2006 07:28:56 UTC