kendallclark: See also RFCs 3402 through 3405/06
evlist: The real nice way to use URIs is to use absolute URIs on the network, but...
evlist: ...networks go down, latency is sometimes significant.
evlist: Catalog based resolution is a solution.
evlist: XML Catalog (OASIS) takes the concept from SGML Open Catalog.
evlist: Caching proxies (such as WWWOffle) are another solution.
evlist: Catalogs and cache work well together, especially if catalog delegation is used.
evlist: ...networks go down, latency is sometimes significant.
evlist: Catalog based resolution is a solution.
evlist: XML Catalog (OASIS) takes the concept from SGML Open Catalog.
evlist: Caching proxies (such as WWWOffle) are another solution.
evlist: Catalogs and cache work well together, especially if catalog delegation is used.
evlist: Namespaces have a significant cost. Benefit is to avoid ambiguity, especially valuable when boundaries are unpredictable.
kendallclark: "I don't want to have to choose between namespaces and validation. I want both."
evlist: Independently developed vocabularies have independent schemas.
kendallclark: "I prefer multiple, small, complementary, single paradigm languages to large monolithiic multi-paradigm ones."
kendallclark: That's interesting, since I agree, but I feel rather diff when it comes to programming languages (that is, I like multiparadigm proggie langs).
evlist: Element tree is divided into sections where all elements have the same namespace.
evlist: An element is a section root if its has a different namespace than its parent.
evlist: "NRL only cares about a section's namespace; it does not care about its internal structure (almost)."
evlist: A NRL schema is a set of rules.
evlist: Rules can contain multiple actions.
evlist: You can combine sections for validation through <attach/> elements.
evlist: You can combine sections for validation.
evlist: Attributes are attached by default to their parent elements but may also be considered as attribute sections.
evlist: NRL supports interleaved namespaces (such as XHTML in XSLT transformations).
evlist: Bigger problem: can NRL help processing multi-namespace documents?
DanC_jam: seems relevant to tag issue mixedNamespaceMeaning-13, which the TAG decided to reduce to 3 smaller issues
DanC_jam: though just because you can validate mixed-namespace documents, you don't necessarily grok their meaning
evlist: That's a feature: DSDL focus is on validation.
kendallclark: "I don't want to have to choose between namespaces and validation. I want both."
evlist: Independently developed vocabularies have independent schemas.
kendallclark: "I prefer multiple, small, complementary, single paradigm languages to large monolithiic multi-paradigm ones."
kendallclark: That's interesting, since I agree, but I feel rather diff when it comes to programming languages (that is, I like multiparadigm proggie langs).
evlist: Element tree is divided into sections where all elements have the same namespace.
evlist: An element is a section root if its has a different namespace than its parent.
evlist: "NRL only cares about a section's namespace; it does not care about its internal structure (almost)."
evlist: A NRL schema is a set of rules.
evlist: Rules can contain multiple actions.
evlist: You can combine sections for validation through <attach/> elements.
evlist: You can combine sections for validation.
evlist: Attributes are attached by default to their parent elements but may also be considered as attribute sections.
evlist: NRL supports interleaved namespaces (such as XHTML in XSLT transformations).
evlist: Bigger problem: can NRL help processing multi-namespace documents?
DanC_jam: seems relevant to tag issue mixedNamespaceMeaning-13, which the TAG decided to reduce to 3 smaller issues
DanC_jam: though just because you can validate mixed-namespace documents, you don't necessarily grok their meaning
evlist: That's a feature: DSDL focus is on validation.
evlist: Murata Makoto shows examples of composite (XHTML + MathML, XForms + XHTML2 + user defined, XHTML 2, MathML, SVG, RDF, ...) documents
MDubinko: "Who has read all these documents: XHTML, RDF, SVG, MathML, XForms" -- one person raises hand
MDubinko: by necessity, mixed vocabulary handling needs to handle partial understanding
evlist: The intention of namespaces was to solve the problem.
evlist: We need to be able to use the schemas for the different namespaces without understanding them.
evlist: These namespaces may use different schema languages.
evlist: Special purpose schema languages will continue to appear: on a small scope they are always more convenient than general purpose schema languages.
evlist: This divide & validate paradigm can be implemented as a SAX events dispatcher.
evlist: If everything goes well, DSDL Part 4 should be a final ISO standard next year.
evlist: After namespaces in XML, this is "the second step for a non-monolithic WWW".
MDubinko: "Who has read all these documents: XHTML, RDF, SVG, MathML, XForms" -- one person raises hand
MDubinko: by necessity, mixed vocabulary handling needs to handle partial understanding
evlist: The intention of namespaces was to solve the problem.
evlist: We need to be able to use the schemas for the different namespaces without understanding them.
evlist: These namespaces may use different schema languages.
evlist: Special purpose schema languages will continue to appear: on a small scope they are always more convenient than general purpose schema languages.
evlist: This divide & validate paradigm can be implemented as a SAX events dispatcher.
evlist: If everything goes well, DSDL Part 4 should be a final ISO standard next year.
evlist: After namespaces in XML, this is "the second step for a non-monolithic WWW".
kendallclark: An Onsite XML Deviant (not really pubbed yet, but it's there for a look for the curious)
MarkB: webarch has gone last call
MarkB: webarch has gone last call
evlist: Requirements: free XML editor that supports RELAX NG that he would use.
evlist: Result: nXML mode for Emacs all written in Emacs Lisp (~15 000 lines).
evlist: If you don't use RELAX NG, you can use trang or msv to convert your schemas into RELAX NG.
evlist: Emacs Lisp is slow and better solutions than brute force had to be considered.
evlist: nXML is a text based XML editor as opposed to editors editing XML structures.
evlist: nXML doesn't require a schema, but when a schema is provided it detects invalidity continuously and provides completion.
edd: Complete continuous revalidation impractical, so must find a way of incremental validation.
edd: Change within validated region moves end of validated region to behinning of change, region bdaries recomputed immediately after each change, validation is postponed uintil either editor is idle or editing operations require validation state.
edd: Choosing the validation algo: must be efficient both to store the validation state, and compare it with a stored state.
edd: Derivative based algos satisfy this requirement and are commonly used for RELAX NG validation
edd: XML parsing -- need to be able to start parsing from a position that has a stored validation state. Full XML parsing can only be done sequentyially. Solution is a different devision of responsibility between parser and validator.
edd: Parser responsibilities -- parse prolog as usual, divide instance into tokens, check all XML constraints that are internal to a single token, provide information about each token.
edd: Validator responsibilities -- check that tags balance, check that namespace prefixes are declared, check for duplicate attributes, check that the instance consists of a single elements
edd: Note some features traditionally associated with the parser have been moved to the validator.
edd: Implications of these new responsibilities: we need a richer validation state to include info about open elements and inscope namespaces. This means the parser can start at any position in the instance, provided it is between tokens
edd: Instance parsing is stateless, store validation state only between tokens, any change to internal subset requires complete reparsing.
evlist: As a summary, James mentions that this would not have work for SGML since he is taking advantage of the simplications of XML.
evlist: He insists that he never builds a tree.
evlist: The parsing is done twice (minimal without validation for syntax highlight) and full with validation.
DanC_jam: Clark demos nxml-mode on an OWL/RDF/XML document
DanC_jam: quite relevant, since the XML Schema WG is on record objecting to RDF's syntax. Perhaps the bug is in XML Schema, not RDF.
DanC_jam: Clark loads NCI's cancer ontology, while praying to the demo gods...
DanC_jam: syntax-highlighting the end of the document works even though only 22% is validated so far
logger: See discussion
evlist: Result: nXML mode for Emacs all written in Emacs Lisp (~15 000 lines).
evlist: If you don't use RELAX NG, you can use trang or msv to convert your schemas into RELAX NG.
evlist: Emacs Lisp is slow and better solutions than brute force had to be considered.
evlist: nXML is a text based XML editor as opposed to editors editing XML structures.
evlist: nXML doesn't require a schema, but when a schema is provided it detects invalidity continuously and provides completion.
edd: Complete continuous revalidation impractical, so must find a way of incremental validation.
edd: Change within validated region moves end of validated region to behinning of change, region bdaries recomputed immediately after each change, validation is postponed uintil either editor is idle or editing operations require validation state.
edd: Choosing the validation algo: must be efficient both to store the validation state, and compare it with a stored state.
edd: Derivative based algos satisfy this requirement and are commonly used for RELAX NG validation
edd: XML parsing -- need to be able to start parsing from a position that has a stored validation state. Full XML parsing can only be done sequentyially. Solution is a different devision of responsibility between parser and validator.
edd: Parser responsibilities -- parse prolog as usual, divide instance into tokens, check all XML constraints that are internal to a single token, provide information about each token.
edd: Validator responsibilities -- check that tags balance, check that namespace prefixes are declared, check for duplicate attributes, check that the instance consists of a single elements
edd: Note some features traditionally associated with the parser have been moved to the validator.
edd: Implications of these new responsibilities: we need a richer validation state to include info about open elements and inscope namespaces. This means the parser can start at any position in the instance, provided it is between tokens
edd: Instance parsing is stateless, store validation state only between tokens, any change to internal subset requires complete reparsing.
evlist: As a summary, James mentions that this would not have work for SGML since he is taking advantage of the simplications of XML.
evlist: He insists that he never builds a tree.
evlist: The parsing is done twice (minimal without validation for syntax highlight) and full with validation.
DanC_jam: Clark demos nxml-mode on an OWL/RDF/XML document
DanC_jam: quite relevant, since the XML Schema WG is on record objecting to RDF's syntax. Perhaps the bug is in XML Schema, not RDF.
DanC_jam: Clark loads NCI's cancer ontology, while praying to the demo gods...
DanC_jam: syntax-highlighting the end of the document works even though only 22% is validated so far
logger: See discussion
kendallclark: Which I'm attending, but not churning -- it's very good.
kendallclark: Mike's said good things about some of the semweb effort w/r/t the WSA work. OWL, for example, has "proven to be quite useful" in the WSA effort. I think this is stuff Bijan Parsia has been doing. At least, Evren Sirin of UMD has been working on it too.
kendallclark: "I've come away with a lot more respect for OWL...than I did a couple of years ago", Mike says.
kendallclark: Yay!
MarkB: some notes of his posted to www-ws-arch
kendallclark: Ah, thanks, MarkB.
kendallclark: He calls NRL a "combinator" schema language; more Haskell influence (at least at the naming level, maybe the algorithmic, too)
kendallclark: (FWIW, in Haskell, a combinator is a function or definition with no free variables.)
kendallclark: Mike's said good things about some of the semweb effort w/r/t the WSA work. OWL, for example, has "proven to be quite useful" in the WSA effort. I think this is stuff Bijan Parsia has been doing. At least, Evren Sirin of UMD has been working on it too.
kendallclark: "I've come away with a lot more respect for OWL...than I did a couple of years ago", Mike says.
kendallclark: Yay!
MarkB: some notes of his posted to www-ws-arch
kendallclark: Ah, thanks, MarkB.
kendallclark: He calls NRL a "combinator" schema language; more Haskell influence (at least at the naming level, maybe the algorithmic, too)
kendallclark: (FWIW, in Haskell, a combinator is a function or definition with no free variables.)
evlist: The plan is to create an active community beyond these DTDs.
evlist: The goal is to preserve the intellectual content of journal articles.
evlist: It's a modular DTD suite that can be used to define custom DTDs.
evlist: Borrows from existing DTDs (MathML, DocBook, ...) when possible.
evlist: The article DTD has been written as a superset of existing DTDs for articles.
evlist: An extensive documentation is available.
evlist: DTDs have been used because W3C XML Schema do cause more interoperability issues and DTDs have been considered easier to modularize.
evlist: But other schema languages are being considered (RELAX NG).
evlist: The adoption by the community is better than they had anticipated.
evlist: The DTDs are not really open source (they want to retain control on the updates) but public domain (can grab them and use them as you like).
evlist: More info at http://dtd.nlm.nih.gov/
evlist: The goal is to preserve the intellectual content of journal articles.
evlist: It's a modular DTD suite that can be used to define custom DTDs.
evlist: Borrows from existing DTDs (MathML, DocBook, ...) when possible.
evlist: The article DTD has been written as a superset of existing DTDs for articles.
evlist: An extensive documentation is available.
evlist: DTDs have been used because W3C XML Schema do cause more interoperability issues and DTDs have been considered easier to modularize.
evlist: But other schema languages are being considered (RELAX NG).
evlist: The adoption by the community is better than they had anticipated.
evlist: The DTDs are not really open source (they want to retain control on the updates) but public domain (can grab them and use them as you like).
evlist: More info at http://dtd.nlm.nih.gov/
danbri: This dropped off the Web; I just dug it out via archive.org.
danbri: Adam Bosworth presented similar position at QL'98 in Boston; see associated paper by Andrew Layman (then a member of the RDFS WG btw).
danbri: See Aug 1998 RDF Schema Working Draft.
danbri: Adam Bosworth presented similar position at QL'98 in Boston; see associated paper by Andrew Layman (then a member of the RDFS WG btw).
danbri: See Aug 1998 RDF Schema Working Draft.
kendallclark: (in his XML Conf, 2003, keynote)
kendallclark: "Programmers don't think about things like SAX and DOM, but rather things like 'purchase order'." -- Huh?
kendallclark: That's rather a bizarre claim, I think.
kendallclark: "There's not a single standard 'out there' which says 'call that service before this one'."
kendallclark: He's seriously slagging really complex web services; sounds like a REST evangelist, though I'm not sure he means to sound like one.
kendallclark: This is actually quite interesting, since BEA is as responsible as any other institution in complexifying web services, particularly the whole EJB thing, which he just slagged hard.
kendallclark: "My service-provider customers want a 'soap cookie', not fancy coordination/transaction stuff."
kendallclark: Pushing hard on "synchronization"; suggests SyncML needs updating.
kendallclark: argh, he pimps weblogs ("there are these things called blog readers") hard, which is bizarre.
kendallclark: "Programmers don't think about things like SAX and DOM, but rather things like 'purchase order'." -- Huh?
kendallclark: That's rather a bizarre claim, I think.
kendallclark: "There's not a single standard 'out there' which says 'call that service before this one'."
kendallclark: He's seriously slagging really complex web services; sounds like a REST evangelist, though I'm not sure he means to sound like one.
kendallclark: This is actually quite interesting, since BEA is as responsible as any other institution in complexifying web services, particularly the whole EJB thing, which he just slagged hard.
kendallclark: "My service-provider customers want a 'soap cookie', not fancy coordination/transaction stuff."
kendallclark: Pushing hard on "synchronization"; suggests SyncML needs updating.
kendallclark: argh, he pimps weblogs ("there are these things called blog readers") hard, which is bizarre.
dajobe: their blog roll isn't growing very fast
DanC_jam: has anybody asked that the rdfig blog be added?
DanC_jam: I just sent a request; to appear in www-archive
DanC_jam: has anybody asked that the rdfig blog be added?
DanC_jam: I just sent a request; to appear in www-archive