See also: IRC log
Previous: 2005-11-15 http://www.w3.org/2005/11/15-vmtf-minutes
<scribe> ACTION: Ralph confirm either 22 Nov or 6 Dec with Matthieu and Ted or propose an alternative [recorded in http://www.w3.org/2005/11/15-vmtf-minutes.html#action01] [DONE]
<scribe> ACTION: [PENDING] Ralph check configurations on W3C site [recorded in http://www.w3.org/2005/11/15-vmtf-minutes.html#action04]
<scribe> ACTION: [DONE] Tom to prepare excerpts from last telecon for inclusion in new Editor's Draft [recorded in http://www.w3.org/2005/11/15-vmtf-minutes.html#action06]
<scribe> ACTION: [DONE] Alistair to move draft into VM space [recorded in http://www.w3.org/2005/11/15-vmtf-minutes.html#action05]
<scribe> ACTION: [DONE] Alistair to add purl.org case to draft note [recorded in http://www.w3.org/2005/11/15-vmtf-minutes.html#action03]
<scribe> ACTION: [DONE] Alistair investigate PURL use case and add to configuration options [recorded in http://www.w3.org/2005/11/15-vmtf-minutes.html#action02]
-> http://lists.w3.org/Archives/Public/public-swbp-wg/2005Nov/0122.html VM Task Force progress report [Tom 2005-11-22]
<danbri_> Summary looks good. Tom's report has "provenance is supported by using the final URI from the chain of redirects as the name of the graph; different URIs represent different versions of a vocabulary." In 'issues under discussion'. Two deployment models, one: put resource in as named graph. Second model: redirect thru generated ID. SPARQL people did not want to force this issue. Looking for reference: http://www.w3.org/TR/rdf-sparql-query/#rdfDataset -- can't find the exact point I want to make.
-> http://www.w3.org/2001/sw/BestPractices/VM/http-examples/2005-11-18/ Configuring Apache HTTP Server for RDFS/OWL Ontologies Cookbook
Alistair: The requirements section of the cookbook could use more attention. Peter Patel-Schneider disagreed with some of the requirements as stated and proposed softer ones. See http://lists.w3.org/Archives/Public/public-swbp-wg/2005Nov/0125.html
-> http://lists.w3.org/Archives/Public/public-swbp-wg/2005Nov/0074.html agenda for discussion purposes
-> http://www.w3.org/2001/sw/BestPractices/VM/http-examples/2005-11-18/#requirements Requirements
<danbri_> (I'd like an apache config eg that allows the administrator to decide easily which version is the default)
Alistair: I revised the requirements to say that RDF should be served by default and make the client specifically request HTML
<Zakim> danbri_, you wanted to note the OWL DL vs Full issue getting more urgent (ISWC feedback) and to say i prefer defaulting to the xhtml
<vivien> http://www.w3.org/2001/sw/BestPractices/VM/http-examples/2005-11-18/#recipe1
http://isegserv.itd.rl.ac.uk/VM/http-examples/#Example1
http://www.w3.org/2001/sw/BestPractices/VM/http-examples/2005-11-18/#recipe1
Alistair: example 1 in the two versions should be the same
<danbri_> (wouldn't a plain ALias directive do the job here?)
Vivien: another alternative is to set the .rdf named file as the directory index
<danbri_> (rewrite module is overkill?)
Alistair: directory index technique only works for "/" namespaces
DanBri: it looked like aliases
would work; using rewrite may be overkill
... but perhaps I'm prematurely optimizing and symmetry across
examples is better
Ralph: unless there is a significant performance hit, I'd go with using the same kind of recipe for all cases
DanBri: performance hit may only be in extreme cases
Vivien: use of rewrite should not be much of a performance hit
<tbaker> example 3: Slash Static Configuration, Minimal
<danbri_> http://bignosebird.com/apache/a9.shtml "There is a performance penalty for placing RewriteEngine directives in your .htaccess file, but I recommend doing so for the following reasons."[..]
Alistair: I changed the configuration between the old and new drafts
-> http://isegserv.itd.rl.ac.uk/VM/http-examples/#Example3 old
Alistair: index.rdf rewrites to
the .rdf
... I discovered that this implementation creates a
circularity
... leads to namespace/index.rdf but namespace/* matches
another patter so there was a special rewrite rule to break
out
... I simplified this in the new draft
<vivien> http://www.w3.org/2001/sw/BestPractices/VM/http-examples/2005-11-18/#recipe2
<vivien> http://httpd.apache.org/docs/1.3/mod/mod_dir.html#directoryindex
Vivien: it may be easier to use DirectoryIndex to replace the last RewriteRule
Alistair: if we redirect
example2/x back to example2/ will this interfere?
... the 303 redirect is there to comply with the TAG resolution
of httpRange-14
-> http://isegserv.itd.rl.ac.uk/VM/http-examples/#Example2 old text
-> http://www.w3.org/2001/sw/BestPractices/VM/http-examples/2005-11-18/#recipe3 new text
Alistair: I changed this recipe to reduce the number of directory levels
Vivien: I wondered why this
recipe did not use built-in content negotiation
... e.g. given both index.xml and index.rdf apache can be
configured to select based on client Accept:
Alistair: we don't really need a
redirect for the HTML content
... but for the RDF content we want to redirect to a version or
snapshot of the ontology
... you can use this new URI in, e.g. SPARQL queries, to ask
for properties of this version of the ontology
Vivien: this is similar to the
W3C Technical Report Latest Version/ This Version
redirects
... on the W3C servers we use symlinks for the Latest
Version
<vivien> RewriteRule ^example3.rdf$ example3-content/2005-10-31.rdf [R=303]
<vivien> RewriteRule ^example3.html$ example3-content/2005-10-31.html [R=303]
Vivien: with rewrite rules like these you can use apache content negotiation
Alistair: is this enough to trigger the automatic content negotiation?
Vivien: I think so
<vivien> RewriteRule ^example3(\.rdf)?$ example3-content/2005-10-31.rdf [R=303]
Alistair: there's a possible alternative configuration, then, for recipe 3
Vivien: this rewrite rule avoids
symlinks, as you can only use symlinks if you have access to
the filesystem
... folks with FTP-only access can't use symlinks
-> http://isegserv.itd.rl.ac.uk/VM/http-examples/#Example4 old
<vivien> RewriteRule ^example4/(.*) example4-content/2005-10-31.html#$1 [R=303,NE]
-> http://www.w3.org/2001/sw/BestPractices/VM/http-examples/2005-11-18/#recipe4 new text
Vivien: example4/ will go to .html#
<vivien> RewriteRule ^example4/ example4-content/2005-10-31.html [R=303,NE]
Vivien: the pattern should probably be (.+)
Alistair: is # followed by nothing a problem? if not, this saves a rule
Vivien: not sure, though certainly browsers will go to the top of the document
<vivien> RewriteRule ^example4/(.+) example4-content/2005-10-31.html#$1 [R=303,NE]
Ralph: whether or not it's legal
to have nothing after '#', it's clearly a different URI from
the no-# one
... this might or might not be important to us
Vivien: still worth considering whether built-in content negotiation can be used here
Alistair: can't see how to make built-in negotiation work with the $1 substitution
-> http://isegserv.itd.rl.ac.uk/VM/http-examples/#Example5 old text
-> http://www.w3.org/2001/sw/BestPractices/VM/http-examples/2005-11-18/#recipe5 new text
Alistair: this changed between old and new. New has more directives. To avoid having to assume that index and multiviews are enabled for the documentation directory
Ralph: will the new version still work if MultiViews is on?
Vivien: since you are using explicit extensions in the new version it shouldn't matter if MultiViews is on or off
Alistair: I turned off MultiViews
because I saw some funny behavior in the rewriting
... can't remember the details right now
... because of trailing '/' on some of the URIs there was some
automatic rewriting going on
... turning off MultiViews fixed this for me
... is it true that turning off MultiViews in .htaccess will
cascade down to subdirectories?
Vivien: yes
<vivien> http://www.example.com/foo/bar/test/
apache will look for .htaccess in test/ first, then in bar/, then in foo/
<tbaker> RalphS, can we remain on the bridge beyond 15:00?
scribe: each one will overwrite the previous rules
yes, but I have HTML TF call at 15:00
<vivien> RewriteRule ^example5/.* example5-content/2005-10-31.rdf [R=303]
Alistair: the next set of recipes have to do with PURLs
-> http://www.w3.org/2001/sw/BestPractices/VM/http-examples/2005-11-18/#recipe6 text
Ralph: PURL is "Persistent
URL"
... it's a deployed service, using 302 redirects currently,
which is important to one of our major customers
... allows people to publish persistent URIs when they don't
own their own server
<vivien> vivien@han-solo:~$ HEAD -S http://purl.oclc.org/net/swbp-vm/example7/
<vivien> HEAD http://purl.oclc.org/net/swbp-vm/example7/ --> 302 Found
<vivien> HEAD http://isegserv.itd.rl.ac.uk/VM/http-examples/example7.rdf --> 200 OK
Tom: At DCMI, we do not need to register individual PURLs for each new property. http://purl.org/dc/elements/1.1/foobar will redirect to an RDF schema with "#foobar" at the end.
<vivien> http://purl.oclc.org/
<vivien> http://purl.oclc.org/maint/choose_redirect.html
<vivien> [[ For example, if the partial redirect http://purl.foo.com/bar/ exists and is associated with the URL http://your.web.server/your/servers/web/root/ then an attempt to resolve the partial redirect PURL http://purl.foo.com/bar/some/other/stuff.html will resolve to the URL http://your.web.server/your/servers/web/root/some/other/stuff.html
<vivien> ]]
<tbaker> ACTION: alistair to revise 7,9,10 for partial redirects [recorded in http://www.w3.org/2005/11/22-vmtf-minutes.html#action07] -- everything Vivien said for 4 and 5 we can take on board for 9 and 10