JonathanRees/Notes
AWWSW
Goal: Develop a formal model (ontology) of the phenomena around HTTP interaction, using RDF. Employ the model to explain as many web architecture principles as possible.
Activities
- Fundamentals - e.g. http://lists.w3.org/Archives/Public/public-awwsw/2008Feb/0019
- N.b. AWWSW has no need to know what a "resource" is
- Ontology
- Alan: Eventually wants a rigorous ontology following methodology of BFO ("basic formal ontology" of Barry Smith); fit with OBO 'Denrie' activity
- JAR: Concerned about persistence, alternative resolution, "authority"
- Tim, David Booth, Noah: "classical" semweb approach
- Rules - Tim, David Booth
domain | property (predicate) |
a thing | is denoted by |
an HTTP response | was / will be / might be received for a GET of |
an HTTP response | carries |
a 'potential representation' | was / will be / might be retrieved by a GET of |
a thing (IR, 'presentation') | had / will have / might have as a 'representation' |
a 'potential representation' | specifies that content is to be governed by |
a 'potential representation' | asserts (says, states) |
Sample questions:
- How do the various 30xs interact with one another?
- Under what circumstances is it valid to infer that what you GOT is a representation of the IR or v.v.?
- Under what circumstances (if any) is it valid to conclude that two URIs name the same IR?
No single answer. A family (or menu) of models depending on which inference rules are admitted.
httpRedirections-57
- FindingResourceDescriptions
- Dozens of potential applications (follow your nose, client simplification, stability policy, access control, bibliographic info, rdf:type, abstract, change log, alternative versions, available variants, site metadata, etc)
- Implemented (Apache configuration & Tabulator consumption)
- Henry: using a response header like this is "egregious"
- Patrick Stickler (URIQA): don't use GET, use MGET, which can be implemented using Apache 500 handler
- Not worth doing without TAG's "fare ye well" & a path to endorsement
- JAR next steps:
- Collect list of interested parties
- Gather requirements / desiderata
- How would TAG like to be involved?
- suggest requirements up front
- ongoing kibbutzing
- occasional reviews
HCLS URI note
- Began in an effort to answer the question "how can we present http: patterns of use that are so good that no one feels the need to use urn:lsid:"?
- Difficult
- Issue is not just technical adequacy but also trust and marketing
- Evolved into a ramble
- Battle lost on the TDWG front
- Battle lost on INCHI front
- Archival quality and so-called "location independence" are the main bugaboos
Rewrite planned, soon, with Michel Dumontier
Interesting challenge to webarch: http://webcitation.org/
What is BMC supposed to do?
- Refuse to accept URIs
- violates 'everything should be named by a URI' principle
- Eggs in one basket, guard the basket
- similar to above since few URIs are archival
- Give the resource a new archival name (and place) (e.g. webcitation.org/...)
- this happens in astronomy
- violates no-aliases principle
- defeats search and semweb joins
- Admit alternative resolution strategies
- precedent: traditional scholarship (e.g. Linnaean system)
- in conflict with webarch notion of 'authority'