rotating an idea in my head re: the classic problem of “overloaded uris” where you don’t know if a uri refers to a document or to the thing being described by the document
-
rotating an idea in my head re: the classic problem of “overloaded uris” where you don’t know if a uri refers to a document or to the thing being described by the document
what if query component
typically /foo returns 303 See Other
- /foo.ttl
- /foo.jsonld
- etcbut what if
- /foo is the thing
- /foo?descriptor is the rdf, which itself responds to content negotiation
— /foo?descriptor.ttl
— /foo?descriptor.jsonldno more “hash vs slash” debates, a new challenger has appeared
-
trwnh@mastodon.socialreplied to trwnh@mastodon.social last edited by
(i am not super serious about this, there may well be problems with this and maybe you shouldn’t actually do this)
(im thinking it could incur +2 roundtrips because you get 303 See Other -> 300 Multiple Choices -> 200 OK) -
trwnh@mastodon.socialreplied to trwnh@mastodon.social last edited by
slightly more level headed thing to do is like. /foo/ and then something like apache conneg for /foo/index.whatever
-
xrisk@social.treehouse.systemsreplied to trwnh@mastodon.social last edited by
@trwnh isn’t the accepts header designed for this purpose? Or am I misunderstanding what you said?
-
trwnh@mastodon.socialreplied to xrisk@social.treehouse.systems last edited by
@xrisk before you get to that point, you might first ask for a Thing and then the server tells you "sorry, i can't serve you the thing itself, i can only serve you something related to the thing like a descriptor document" (this is what 303 See Other does)
then you ask for the Other and the server tells you "oh i have this Other, but in multiple formats" (this is what 300 Multiple Choices does)
using an HTTP Accept header lets you skip the second step but not the first