capturing control-c: http://stackoverflow.com/questions/2089421/capturing-ctrl-c-in-ruby begin do stuff clear bookmark rescue Interrupt write bookmark fall through (with notation) end print report pairtree http://www.rubydoc.info/gems/pairtree/0.1.0/ What should the filenames be? Does it matter? sinatra http://www.sinatrarb.com/intro.html http://www.kalzumeus.com/2010/01/15/deploying-sinatra-on-ubuntu-in-which-i-employ-a-secretary/ serializing: http://www.sitepoint.com/choosing-right-serialization-format/ content negotiation: http://www.w3.org/TR/swbp-vocab-pub/#recipe3
Responses: Home page: something helpful and descriptive termsOfUse: describe the license URI: If the individual exists, look at the accept header, and forward appropriately. default is rdf/xml Otherwise, 404 with an informative message. URI/localname.* If the type is not supported, 404 with an informative message If the URI can't be determined, 404 with an informative message If the individual doesn't exist, 404 with an informative message Other 404 with an informative message
Do we honor a request for HTML? What should we show? Do we need to worry about the name of the file that is served, or does the browser worry about it? appending /localname.type means that this will be used as the filename, in most browsers. The server must add the doc-specific triples, if only because the name of the doc depends on the format get the URI of the doc from the file properties get the label from the label of the subject (default to something if no label) get the date from the file properties publisher is the base of the URI rights points to termsOfUse page.
http://wifo5-03.informatik.uni-mannheim.de/bizer/pub/LinkedDataTutorial/#deref
http://vivo.cornell.edu/individual/CRW-dre39/CRW-dre39.n3 a foaf:Document ; rdfs:label "RDF description of Egerton, D - http://vivo.cornell.edu/individual/CRW-dre39" ; dc:date "2015-11-11T11:23:40"^^xsd:dateTime ; dc:publisher http://vivo.cornell.edu ; dc:rights http://vivo.cornell.edu/termsOfUse .
More appropriate queries: titles as well as labels, etc. Look to the indexer for these? Apply different queries to different classes of object?
The final namespaces will be something like: http://draft.ld4l.org/cornell/ http://draft.ld4l.org/harvard/ http://draft.ld4l.org/stanford/
Three instances of the app, three pairtree systems, pass args on startup Pairtree: prefix is http://draft.ld4l.org/cornell/, so only the localname In config.ru: run AwesomeApp.new(app, bootstrap: true) http://stackoverflow.com/questions/9657202/pass-arguments-to-new-sinatra-app
map('/example') { run ExampleController }
map('/') { run ApplicationController }
https://www.safaribooksonline.com/library/view/sinatra-up-and/9781449306847/ch04.html
But then how do we create the three pairtree systems?
Maybe one pairtree, prefix is http://draft.ld4l.org/ and the app uses a parameter to know its context path
The nice thing about three apps is that we can still redirect based on localname (can we? Or do we need an absolute redirect mechanism?) Needs to be absolute but we can do that.
The prefix approach Generator works as now, but with a required prefix as an argument. Only generates files for URIs that use the prefix Give it http://draft.ld4l.org/, and it will take it from there: Point out: this is not a namespace, this is a prefix. One pairtree, with paths that look like: co/rn/el/l-/00/00/... App can determine context from URI, but should it have to? The namespace approach: Generator works with namespace as an argument. Must be run three times with three different namespaces to create three pairtrees paths look like 00/00/... Three apps serving the three pairtrees. How do we do redirects? We still get the full URL Mix and match: Generator works with the prefix approach. 3 servers with the namespace approach.