Re: RDF(-DEV), back to the future (was Re: Semantic Web Interest Group now closed)

tl;dr -- +1 to rdf-dev as a first-class entity

Many thanks for this detailed history, Dan. As someone who first tuned in
to the Semantic Web in 2006, I am always fascinated to see how much deeper
the rabbit hole goes, and how nonlinear the progression from semantics +
Web to the Semantic Web. In my own much more limited experience, the
intricate history of a technology is always oversimplified in the minds of
newcomers to the field; the Velvet Undergrounds of the world simply don't
get the kind of broad recognition that the Beatles do, even if the former
can be said to have enabled the latter.

Since you mention property graphs, it may be worth pointing out a Beatles /
Velvet Underground relationship there, as well; it is not generally
appreciated that property graphs owe a lot of their current popularity to
RDF. From what I have seen in industry, newcomers to the graph database
community tend to view RDF and SPARQL as exotic beasts somehow connected to
"knowledge graphs". However, TinkerPop and Gremlin were originally a
reaction to RDF (described above as "painfully simple"!) coupled with the
excitement of the Linked Data movement and the availability of new,
triple-store-like databases including, first and foremost, Neo4j. These
tools gave developers much of the expressive power of RDF without the
apparent baggage of namespaces, blank nodes and other doodads, and the
association with complex serialization formats and intimidating ontology
frameworks. To Henry's point about pragmatics, property graphs scored with
developers by appearing to get out of the way and just let them get stuff
done. To this day, there isn't even a formal specification of property
graphs, despite a 2013 effort by members of the SemWeb community to create
one. That said, there is a huge need these days for a little more
conceptual rigor in the graph database community, and RDF is helping to
fill that gap. See me butcher some of the history here:
https://drive.google.com/file/d/0B9GF5nhNxdyvc2ZMMkJSRF9IcUlmeE04Qm1KX291ejI2V2pz/view

There is another gap to bridge between RDF and property graphs on the one
hand, and algebraic data types on the other -- not only in a relational
context, but also in the context of data interchange frameworks such as
Protocol Buffers, Apache Thrift, Apache Avro, etc. which are close to the
source of truth for a lot of today's corporate big data. Compatibility with
RDF has a lot of still-untapped potential for extending these modern webs
of data beyond company walls, and there is plenty of related work to point
the way. Henry's focus on category theory and RDF is particularly timely,
IMO, in terms of community engagement. Both RDF and Property Graphs are
data models with a basis in set theory, of which category theory is seen by
many as a younger and hipper cousin. A closer correspondence of RDF with
category theory is likely to make it relevant to developers who do not
immediately appreciate its history or deep support for composition of
schemas and data sets.

Josh


On Tue, Oct 16, 2018 at 12:58 AM Henry Story <henry.story@bblfish.net>
wrote:

> Thanks Dan for the detailed story on the origins of RDF. (I am cc-ing it
> over to the
> philosophy community group as origins and philosophy have much in common)
>
> The split we all notice between pragmatics (doing things) and description
> of things is a
> deep philosophical split that according to Robert Brandom goes back to
> Descartes and Hume.
> It has split philosophy in many ways with many attempts to bridge the
> divide. A divide so
> deep is no accident, but reflects a conceptual reality.  And so one
> should  perhaps try to
> confront the problem head on.
>
> As you know I have been always working on build useable apps with RDF,
> coming therefore
> from the Linked Data space before it was called that, having gone to the
> 2004 FOAF Galway
> workshop.  But as I built things I discovered that engineers themselves had
> philosophical/conceptual misunderstandings that since they were not voiced
> and discussed,
> being too far back in the origins, concealed as it were from view, kept
> pushing
> engineering projects off the rails.
>
> The title Semantic Web has this contradiction in its name. But I would
> like to argue that
> that is a good thing, as it names the bridge that needs to be crossed,
> between pragmatics
> and analytics.  I mentioned Robert Brandom because having written a book
> on "The Logic of
> Inconsistency" in 1979, has then become famous for works such as "Between
> Saying and
> Doing: Towards an Analytic Pragmatism".  Brandom was a student of David
> Lewis (he clearly
> looks like Lewis, but he rarely mentions him), and a big fan of Rorty and
> most of all the
> American Pragmatist Willfred Sellars. He then succeeds with an amazing
> feat for a
> philosopher coming from the analytic tradition, which is to make Hegel
> relevant again.
> Hegel let us remember is the philosopher of history as conceptual
> changes.  of thesis,
> antithesis and synthesis.
>
> I think we are actually reaching a synthesis point, and Brandom shows the
> way.  His aim is
> to show how pragmatics underlies semantics.  To say that someone believes
> something one
> must have some criteria about what they need to do for that to be
> attributable to them.
> What they need to do is to be committed to the consequences of the
> statement, reject the
> incoherences, and in a way update their belief (store). Modal statements
> are just
> explicitations he argues of what we already do when we  use everyday
> language.  To state
> that placing the cup on the table it won't shatter, is to know that there
> are exceptions
> to that expressed in subjunctive modes relating to exceptions such as the
> table not being
> pushed, or someone throwing a rock at the cup,...  We as hyper-data
> application writers
> are not dealing directly with natural language so we may be able to reduce
> the complexity
> of counterfactual reasoning, by working as engineers within the
> constraints of the
> reasoning engines available, and offloading counterfacual reasoning onto
> humans that use
> the applications we build.  ( I gave an intro to this way of thinking at
> ISWC
> https://twitter.com/bblfish/status/1049343868569837568 ) But the idea
> that we should think
> of how vocabularies are used and that that underlies the semantics is I
> think the right
> way to think of the project here.
>
> As it happens we had a very interesting conversation on this mailing list
> just last month
> regarding category theory, which is now used by programmers all the time -
> but used to be
> laughed at as abstract nonsense 20 years ago! Clearly nonsense is an
> important aspect of
> the life, one that should not be overlook, and that is also very effective
> in a chaotic
> way in the real world. Note: There are semantics of impossibilities that
> have been developed
> over the past 40 years, and that are now used to explain how imagination
> works.
> https://twitter.com/bblfish/status/1051919789503975428
>
>
> But more to the point category theory is the mathematics of symmetries.
> One important
> symmetry is between that of algebras -- the high school sort: addition,
> multiplication,
> subtraction,... but also logic, geometry, topology -- and coalgebras, the
> infinite
> streams, processes etc.. This symmetry looks to me to be very similar to
> the symmetry
> between saying and doing. Indeed in the 2014 lecture "Analytic Pragmatism,
> Expressivism
> and Modality" [1] Brandom uses an example of a Father Christmas machine
> that can produce
> and recognize sentences such as ‘hahaha!’, ‘hohoho!’, ‘hahahoho!’
> ‘hohoha!’. Interestingly
> he points out languages to describe such machines can be less expressive,
> ie lower
> down in the Chomsky Hierarchy: as we all know c++, Php and most languages
> used to program
> turing machines are much less expressive than those turing machines. So
> that one can
> bootstrap more complex languages and processes with simpler ones. Now: the
> Father Chrismtas
> machine is coalgebraic and the languages used to program them are
> algebraic. Both work
> together.
>
> So I think we are reaching a point where we can mathamtically and
> philosophically bridge
> the divide between pragmatic linked data and semantic modelling. It seems
> therefore that
> closing the list right now would be just the thing that the Hitch Hikers
> Guide to the
> Galaxy predicted always happens just at the point when the meaning of
> something important
> is about to be discovered.
>
> If the community groups have more facilities then I would suggest just
> creating one
> there without changing the mailing list. Just keeping everything as is.
> Perhaps one
> could emphasize in the mission statement the tension between pragmatics
> and semantics
> that is built into the term semantic web, and that discussions that try to
> resolve
> those issues are particualarly welcome.
>
>
> All the best,
>
>         Henry Story
>
>
> [1] http://www.pitt.edu/~brandom/currentwork.html
>
>
>
>
> > On 16 Oct 2018, at 04:53, Dan Brickley <danbri@google.com> wrote:
> >
> >
> > This is way overly long, sorry. I wrote down some pre-2001 RDF history,
> since this community transition point is a good time to take stock and
> remember SWIG/RDFIG and RDF-DEV things that we forgot to record, before it
> is too late.
> >
> > On Mon, 15 Oct 2018 at 16:06, Melvin Carvalho <melvincarvalho@gmail.com>
> wrote:
> >
> >
> > On Tue, 16 Oct 2018 at 00:53, Dan Brickley <danbri@google.com> wrote:
> >
> >
> > On Mon, 15 Oct 2018, 12:32 Ralph Swick, <swick@w3.org> wrote:
> >
> > On 2018-10-15 11:09 AM, David Booth wrote:
> > > On 10/15/2018 10:49 AM, xueyuan wrote:
> > >  > This message is to inform you that the Semantic Web Interest Group
> > >  > is now closed, [ . . . . ]
> > >  > With the introduction of Community Groups we now encourage the
> > >  > participants in the IG forum to
> > >  > establish Community Groups to continue the conversations.
> > >
> > > Given that the semantic-web@w3.org email list has served the
> community
> > > very well, I think it would be helpful for continuity if a Community
> > > Group could take over the existing email list.  Is this possible?  And
> > > if so, does this mean that we should now create such a community group?
> >
> > Ivan and I have been in conversation with DanBri for some time as the
> > formal closing of the Interest Group was pending.  This specific
> > question was part of that discussion; whether to continue the big
> > semantic-web distribution list as a Community Group resource or use the
> > opportunity to do some housekeeping.
> >
> > Ivan and I decided to let the community decide -- and those discussions
> > are welcome on the list.
> >
> > And again, I can't overstate our appreciate to DanBri for his gentle
> > facilitation of the discussions on this list, jumping in as the IG chair
> > and list moderator only when it was critical to do so.
> >
> > Thanks Ralph. I had hoped to propose a new followup Community Group last
> week but got swept up in f2f discussions during the ISWC conference.
> >
> > Both SW and Linked Data have rather prescriptive overtones (1-star,
> 5-star, #-/ redirects etc.). My suggestion to Ralph, Ivan and team was to
> go back to the original name we used prior to creation of 1999's RDF
> Interest Group. It was "RDF-DEV" originally, named in tribute to XML's now
> decades-spanning XML-DEV community.
> >
> > Linked data already has a list.
> >
> > I think changing the name of something that's been going a fair requires
> some onus of the proposer to justify it.
> >
> > We can potentially keep both semantic-web@ and public-lod@ but we
> cannot take their existing membership and treat those unknown parties as
> W3C Community Group members; the CG process doesn't work like that, since
> there are various things to agree to when joining a CG. So the suggest is a
> new lightweight rdf-dev@ Community Group, even with most discussions
> staying at least for now on their current lists.
> >
> > Regarding the specific motivation, it would be good to look at.
> >
> > Prescriptive.  Not sure what this alludes to.  There have been debates
> over different quality of data (1 star - 5 star) but surely that is not
> only as expected, but as designed!
> >
> > (The "1-5 star" viewpoint is one of several views on how to deploy RDF
> at scale. There are other perspectives that shift burdens from publishers
> to consumers, neither is wrong or right, just a landscape of tradeoffs and
> compromises.)
> >
> > The semantic web gives you a protocol where one set of data can
> interface with another.  So the degree of plumbing goes from the network,
> to the data.  Instead of looking at packets you're looking at data shapes.
> So isnt it only natural that data quality becomes an increasing topic of
> interest.
> >
> > On the specific case of #-/ redirects, tatooed agents not withstanding,
> this is simply a conversation about data shapes, isnt it (maybe im using
> the wrong word there)?  In some systems the data model overloads the shape
> of data so that a URI points to a document and class.  This for some is a
> neat slight of hand, and no future analysis is needed.  For others the
> overloading causes edge cases which are hard to resolve.  The example I
> once gave is, "I might like RIcky Martin's home page, but I might not lick
> RIcky Martin".  Isn't this the kind of discussion that is to be encouraged
> as we start to learn to put data together, and learn about interop?
> >
> > Final observation.  I came to this community as a skeptic.  For many the
> term "rdf" doesnt mean much, but the term "semantic web" is magic.
> Outsiders dont know what it does, they know it's complex, too complex for
> them, but they also know it contains a dark power, that if one day is
> unleashed, will be a game changer.  I think it's a mixed brand but a
> powerful one.  Not heard enough yet to feel like ditching it, but am open
> and interested.
> >
> > (A few more general thoughts on SWIG/RDFIG and staying engaged with our
> historical roots, now that I am typing on a computer with a keyboard)
> >
> > Firstly to respond on the naming point: not using a slogan as a
> Community Group name is absolutely not the same as ditching it! We will
> have many slogans, for many contexts. For better or worse, "Linked Data"
> has come to be associated with certain very specific notions of publication
> best practice for *public* RDF data, and "Semantic Web" has acquired
> different overtones (("of dark powers? :)). There is a common element to
> both, and that is the idea of having a shared structured data model based
> around a graph abstraction. As a shorthand for that idea in W3C circles, I
> think it's fair to just say "RDF".
> >
> >
> >
> > For me personally, "Semantic Web"  (https://www.w3.org/1999/11/SW/
> https://www.w3.org/Talks/2001/12-semweb-offices/all.htm ... ) was always
> and remains a project to improve the Web. You can see bits of that history
> in the old public-but-obscure sw99 list,
> https://lists.w3.org/Archives/Public/sw99/1999OctDec/ and in the
> research-funded efforts that the W3C Semantic Web team ran via MIT (
> https://www.w3.org/2000/01/sw/) and in Europe via INRIA/ERCIM,
> https://www.w3.org/2001/sw/Europe/
> >
> > Over time (2001-5ish?) some of us who had been building things around
> practical applications (RSS, Dublin Core, Annotea, FOAF, Mozilla, crawling,
> sitemaps, MCF etc.) began to feel a bit marginalized by the direction of
> the "Semantic Web" slogan, in that conferences etc were organizing around
> the notion of "Semantic Web as a scientific research field" rather than as
> a shared endeavour to improve the Web. Simple and useful applications began
> to look insufficiently researchy from a scholarly perspective, and the
> later "Semantic Web" community tended towards a fixation on
> rules/inference/ontologies as the centerpiece technology rather than as a
> means to an end. At the same time the influx of new participants certainly
> helped in lots of other ways, bringing rigour (sometimes too much rigour:)
> to specs, adding long-anticipated extras to RDF for ontology specification,
> etc. Meanwhile, the approaches to hypertext linked RDF that some of us were
> exploring around FOAF (and which borrowed from Mozilla/MCF ideas) was
> crystalized with TimBL's Linked Data note and - boosted by the idea of
> focussing on open data - https://www.w3.org/DesignIssues/LinkedData -
> that effort became something of a second hub of W3C RDF-related activity,
> for those whose interests were not primarily in the logic/reasoning arena.
> Several practically minded RDF implementors de-camped to the public-lod@
> mailing list, the community became rather divided into two clusters, and to
> some extent at least, "Semantic Web" has drifted into being more of a
> topical research area than a shared project to improve the Web. W3C stopped
> using the "Semantic Web" name actively for new WGs some years ago. Most
> recently there have been discussions about how graph databases (property
> graphs, gremlin etc.) relate to W3C's existing portfolio of specs, and some
> engagement with the emerging "Knowledge Graphs" perspective. My feeling
> from those conversations, and especially from last week's excellent ISWC,
> is that the different strands of activity are gradually re-converging, with
> a broadly RDF-ish notion of graph data as the shared core, rather than an
> over-riding emphasis on inference/ontology, or on particular (e.g. LOD)
> publication patterns or requirements that all structured data publishers
> must meet. Re-centering on the shared graph data model absolutely doesn't
> preclude initiatives to find stronger consensus, whether they are around
> query, OWL2, SHACL/ShEx, WebID/Solid, or the never-ending quest for the
> perfect syntax. That's where I feel we are today, but I wanted also to
> share some historical notes on how we got here (as a cluster of efforts
> touching W3C via SWIG, RDFIG, RDF-DEV).
> >
> >
> > The first public RDF draft is 21 years old this month. See
> https://www.w3.org/TR/WD-rdf-syntax-971002/
> >
> > I met Ralph Swick and Eric Miller at the Dublin Core conference in
> Finland that week in October 1997, alongside Dublin Core folk like Tom
> Baker and Stu Weibel, having read this first public RDF draft on the plane
> flight out, and asked them whether RDF's schema language would be at least
> as expressive as MCF's, i.e. https://www.w3.org/TR/NOTE-MCF-XML/
> https://www.w3.org/TR/NOTE-MCF-XML/MCF-tutorial.html and
> http://www.guha.com/mcf/wp.html ).  They said that they thought so, and
> I've been involved with the RDF effort in one capacity or another ever
> since. In those days, W3C was patterned loosely after the X Consortium, and
> all Working Groups operated in private, members-only fora. When we started
> the RDF Interest Group in 1999 it built a year or so's early adopter
> discussions on the RDF-DEV list,
> http://web.archive.org/web/19990508122105/www.mailbase.ac.uk/lists/rdf-dev/archive.html
> ... that's the reason I'm proposing the otherwise quirky "RDF-DEV' name
> once more.
> >
> > Not everyone here will remember RDF-DEV, and the archives are currently
> offline except via Archive.org (although I'm trying to fix that). It might
> be a good moment to spend a little time looking back on early RDF, and on
> RDF-DEV, as we take stock of where we stand today.
> >
> > The RDF-DEV list was announced to XML-DEV and beyond back in June of
> 1998,
> http://web.archive.org/web/19991022011435/http://www.mailbase.ac.uk:80/lists/rdf-dev/1998-06/0000.html
> http://lists.xml.org/archives/xml-dev/199806/msg00414.html and our
> special relationship with XML continues to this day. Many of the stars of
> the XML world came eventually to dabble in RDF, and many early RDF and
> Semantic Web discussions occurred in XML fora, notably XML.com (e.g.
> https://www.xml.com/pub/a/2001/07/25/prologrdf.html or
> https://www.xml.com/pub/a/98/06/rdf.html ) and sites like XMLHack.com or
> xmlfr.com, http://xmlfr.org/actualites/tech/001208-0001 . We also wrote
> more in IRC chats, blogs and email than we did for the scholarly record,
> and many of those old sites are bit-rotting away.
> >
> > Looking at the first month's archive for RDF-DEV,
> http://web.archive.org/web/19990508122105/www.mailbase.ac.uk/lists/rdf-dev/archive.html
> ->
> http://web.archive.org/web/19990420005959/http://www.mailbase.ac.uk:80/lists/rdf-dev/1998-06/index.html
> I am happy to say that many of the same folk are still involved (e.g. I saw
> Ron Daniel at ISWC this week, Guha is a colleague at Google and founder of
> Schema.org). It is easy in 2018 to forget how things were in those times.
> Not only was there no Google Chrome, there was barely a Google (
> https://en.wikipedia.org/wiki/History_of_Google#Early_history) .
> >
> > Industry adoption discussions tended back then to centre around what
> Netscape (the big  browser of the early Web years) were doing, and
> speculation or concern about whether Microsoft would engage substantively
> with the RDF effort or not. I have long felt that our tendency in this
> community effort towards repeatedly renaming things (PICS -> PICS-NG -> RDF
> -> Semantic Web / Linked Data, ...) has disconnected us from our own origin
> stories. While "RDF" might be a terrible name, it is our terrible name, and
> it anchors things back in a solid chain of cause-and-effect going back to
> the Web's younger years.
> >
> > Technically, the RDF approach owes a lot (even most) to MCF (
> https://en.wikipedia.org/wiki/Meta_Content_Framework), and I suspect that
> were it not for browser-war era politicking, we might today be talking
> about W3C's 'Meta Content' rather then 'Resource Description' Framework.
> That said, certain ideas were in the air more broadly, and the RDF specs
> had other important ancestors alongside MCF. There was PICS-NG, an effort
> to upgrade the PICS content labelling system to cover more metadata
> usecases (see Ora Lassila's drafts,
> https://www.w3.org/TR/NOTE-pics-ng-metadata and nearby); this was paired
> with efforts around W3C Digital Signature, with expectation that dsig would
> be commonly used with PICS(-NG) "content labels". There were also efforts
> around Dublin Core, which were often more akin to requirements gathering.
> Beyond the initially quite basic DC vocabulary you can see the origin of
> some of our concerns about modularity and decentralization articulated as
> the "Warwick Framework", an outcome of a 1996 Dublin Core workshop
> http://www.dlib.org/dlib/july96/lagoze/07lagoze.html - or see Michael
> Sperberg-McQueen's note -
> http://dublincore.org/documents/2001/03/19/info-factoring/ - on using
> good old-fashioned logic to disentangle the different markup and
> extensibility ideas that were bouncing around in Dublin Core community
> discussions, again in 1996.
> >
> > Also of that era, the era that gave us RDF, is this semi-published
> sketch of a W3C Note on link types -
> https://www.w3.org/Architecture/NOTE-link.html . It references other
> metadata syntaxes of those times, including SOIF (anyone remember SOIF? see
> https://www.w3.org/Search/9605-Indexing-Workshop/Papers/Hardy@Netscape.html
> https://web.archive.org/web/19971221220012/http://harvest.transarc.com/
> etc.). Other formats/protocols included WHOIS++, as well as "IAFA
> Templates", approaches to representing and indexing the contents of FTP
> sites which some of us had begun to use for Web crawling, indexing,
> cataloguing and data sharing -
> https://www.w3.org/Conferences/WWW4/Papers/52/
> http://www.dlib.org/dlib/january98/01kirriemuir.html etc. Before even RDF
> was created, there was a convergence focussed on W3C's Metadata Activity
> that was beginning to bring together some of these metadata efforts,
> combining digital library aspects with mainstream industry adoption - for
> example see Jon Knight's article on Dublin Core and MCF -
> http://www.ariadne.ac.uk/issue7/mcf. Annotations and metadata services
> were another common theme, with Marja-Riitta Koivunen leading W3C's RDF
> prototype-based explorations ("Annotea"), which eventually matured into the
> W3C Annotation standards. PICS itself always included a labelling service
> protocol (https://www.w3.org/PICS/labels.html), as well as a mechanism
> for content-filtering (or selecting) rules expressed against those metadata
> labels. SPARQL was an eventual migration of this idea to the RDF world.
> >
> > There were also pretty early Web discussions on querying RDF, and those
> tended to have one foot in KR, one foot in databases, and a third in
> digital libraries.
> https://web.archive.org/web/19990423130239/http://www.w3.org:80/TandS/QL/QL98/pp.html
> might be of interest to archaeologists, but I won't dig into that topic
> here.
> >
> > Again in terms of ideas being "in the air", the efforts that fed into
> the 1997 RDF initiative ("W3C Metadata Activity", led by Ralph Swick) had
> ancestry in the Web itself, both in TimBL's original pitch within CERN with
> its now famous diagram https://www.w3.org/History/1989/proposal.html and
> talk of nodes and arcs for question answering. The document that gave us
> the Web, presented a very RDF-like diagramming style for linked
> information, and said:
> >
> > "The sort of information we are discussing answers, for example,
> questions like
> >
> >  - Where is this module used?
> > - Who wrote this code? Where does he work?
> > - What documents exist about that concept?
> > - Which laboratories are included in that project?
> > - Which systems depend on this device?
> > - What documents refer to this one?"
> >
> > If you dig a bit further back (~ 9 years - thanks, Sean Palmer) you can
> find TimBL's 1980 manual for the ENQUIRE system, which makes similar
> observations: http://infomesh.net/2001/enquire/manual/#h2
> >
> > "The  ENQUIRE  system  is designed  to fill  a gap  in many  current
> > documentation  systems.    A person finding himself faced with a piece
> > "xxx" of a system should be able to ask ENQUIRE, for example
> >        What is xxx part of?
> >        What is xxx composed of?
> >        What must I alter if I change xxx?
> >        What facilities does xx use?
> >        Where do I find out more about xxx?
> >        Who or what produced xxx?
> >
> > ENQUIRE does not aim to answer such questions as
> >        How does xxx work?
> >        What format is xxx in, exactly?
> >        Why was xxx created?
> >        What is the format of the interface between xxx and yyy?"
> >
> > Other "in the air" aspects of that era included the SHOE work  (
> https://www.cs.umd.edu/projects/plus/SHOE/ and
> http://www.cs.umd.edu/projects/plus/SHOE/piq.html), which explored many
> of these themes and  - again - bridged KR with the specifics of 1990s Web
> technology.
> >
> > RDF worked as a standardization effort, despite its challenges, not
> because it was super sophisticated, but because it was painfully simple. It
> was in 1997/8 also a kind of Rorshach Test -
> https://en.wikipedia.org/wiki/Rorschach_test - in that different
> communities looked at it and saw something rather familiar looking back at
> them. Early internet technologists and hackers looked at it and saw it as
> an evolution of metadata formats and protocols that had already proved
> themselves, and which were suffering growing pains around extensibility and
> modularity. Dublin Core folk saw a potential answer to their need for
> practical answers on how to embed simple catalogue records within Web
> pages, as well as a logic-oriented, mappings-friendly underpinning that
> could help broker collaborative agreements with nearby metadata communities
> around Education, Multimedia, or Rights (
> http://www.dublincore.org/news/2000/2000-12-06_press-release-e-learning-takes-important-step-forward/
> http://www.dlib.org/dlib/january99/bearman/01bearman.html
> https://web.archive.org/web/20000925085014/http://www.ilrt.bris.ac.uk/discovery/harmony/docs/abc/abc_draft.html
> ...). Those more focused on policy aspects of 1990s internet also saw a
> technological tool relevant to their social concerns, whether it was issues
> around censorship, filtering, labelling technologies (
> https://www.w3.org/TR/NOTE-PICS-Statement ...), privacy-oriented efforts
> such as P3P, or the Web accessibility agenda, RDF looked ... compellingly
> relevant.
> >
> > As 1997's RDF matured into the 1999 W3C "Model and Syntax"
> recommendation (https://www.w3.org/TR/1999/REC-rdf-syntax-19990222/ ) it
> also quite naturally attracted interest in the knowledge representation and
> ontology world. Aside from RDF's MCF heritage which drew in turn upon the
> Cyc project, the RDF approach looked familiar enough to everyone working
> with logic-based representations. My officemate Joel Crisp said of the
> first public spec "But that's just Prolog!". He still says that. My later
> officemate Jan Grant explored the point further by making a Javascript
> prolog engine that ran in the browser (before JS was cool),
> https://www.w3.org/1999/11/11-WWWProposal/rdfqdemo.html and which we
> later explored integrating into Mozilla, alongside (thanks to Geoff
> Chappell) the most robust SWI-Prolog engine (
> https://www-archive.mozilla.org/rdf/doc/inference.html). While I am
> namedropping former Bristol colleagues I should also mention Nikki Rogers,
> who wired up RDF data sources into Coral (a deductive database of that
> era), and Libby Miller, who amongst many many other things, built upon a
> 1998 MSc project codebase implementing MCF from Larry Franklin (now
> Permanent Secretary in the Government of Anguilla; sometimes people move on
> from RDF). Libby implemented server-side in her Java RDF codebase
> equivalents to various things that were in the Netscape and Mozilla efforts
> of the time, including several RDF sitemap format experiments which were
> effectively the ancestor to the "RDFWeb" hyper-linked RDF approach that
> evolved (https://www.w3.org/2001/sw/Europe/talks/xml2003/all.htm) into
> FOAF.
> >
> > It is also easy in 2018 to under-estimate the scale, importance and
> impact of the Netscape and Mozilla efforts around RDF. The early RDF-DEV
> archives linked above show discussions around the RDF/XML markup in
> Netscape's  Netcenter portal. Amongst other things their initiatives
> included sitemap formats, the creation of RSS feeds (which gave us RSS1,
> Atom etc.), large scale open data in RDF (DMOZ, the DBPedia/Wikidata of its
> day, ChefMoz, ...). They also gave us a vision of an RDF-capable Web
> browser (https://www-archive.mozilla.org/rdf/doc/), whose stacked API of
> super-imposed RDF data sources inspired many of the discussions around W3C
> on APIs and RDF querying, and nudged us towards a healthy culture for W3C
> RDF work of grounding standards work in technology ideas that were already
> being explored and refined by implementors. For a while Netscape's RDF
> sitemap (dublin core with extras) was served up with my inline comments,
> http://web.archive.org/web/20000816174920/home.netscape.com/netcenter.rdf
> . This format is long gone and dead today, but it was the ancestor of
> modern Linked Data.
> > Oh, Netscape the browser also pinged Netcenter on each new page visit,
> and fetched an RDF/XML description of the page, which was used to power
> their "what's related" UI.
> >
> > Why have I spent my evening digging up these rather long-lost
> experiments, debates and proposals?  These were the kinds of things that we
> discussed in the RDF-DEV list in 1998-9, and the seeds of the W3C Interest
> Group for RDF (RDFIG) which was created back in August 1999
> https://lists.w3.org/Archives/Public/www-rdf-interest/1999Aug/0000.html
> where another horribly long email of mine -
> https://lists.w3.org/Archives/Public/www-rdf-interest/1999Aug/0004.html -
> introduced several themes of the era, including interop testing and the
> (somewhat troubled) relationship with related efforts around XML.
> >
> > I didn't intend this mail to be a huge nostalgia trip or name-dropping
> of friends from the late 1990s. There are many more important contributions
> I've not listed; I won't attempt to name and list them here. It is more
> that I have been realizing lately that the origin myths around Semantic Web
> and Linked Data have disconnected it from its actual historical roots.
> Please let nothing I've said here be taken as a slight to, or undermining
> of, the amazing contributions and important milestones that came after
> 2000/2001. For example, the Scientific American article was a major
> milestone, TimBL's "Linked Data note" and later TED talk was another. We
> have, through a lack of institutional memory and history-writing allowed
> newer members of our community to slip into seeing those milestones as the
> actual origins of the RDF/SW/Linked Data work. This is a mistake that
> disconnects us from how we got here, and can distort our thinking about
> where we go next. RDF is a specific project, created under the umbrella of
> a particular (if sometimes peculiar) organization, and it is a pity if we
> lose track of that project now that it touches three decades. It is a
> project intimately tied up with the larger Web project itself, and one that
> deserves some kind of ongoing bridge between W3C the organization, and W3C
> the wider network of communities of interest who hang around here on
> mailing lists. And it is a project that was substantially underway already
> in 1997-2000, and whose discussions in those earliest years also formed the
> basis for many of the earliest (and most useful) activities around W3C
> RDFIG/SWIG as a group.
> >
> > RDF of course is nothing terribly new as an approach to describing
> things in a KR or logical manner, and many of its common themes and
> challenges have been debated intensely for decades or longer. That is the
> point of standards, to find and specify common ground rather than to invent
> new things in a committee. Thinking of KR and picking from a longer list,
> ... the "What's in a Link" paper (
> http://www.dtic.mil/dtic/tr/fulltext/u2/a022584.pdf) dates from 1975.
> "Shakespeare's Plays Weren't Written by Him, but by Someone Else of the
> Same Name" (https://philpapers.org/rec/HOFSPW) dates from 1982. These and
> others demonstrate a longer, larger KR/AI heritage dating back before the
> Web, the Internet, and computing into the history of Philosophy and logic.
> If you prefer a different heritage, you can travel back in time through the
> library world, to the origins of cataloguing, perhaps the Dewey Decimal and
> its early fork, the UDC classification, or the Mundaneum (1912's idea of a
> search engine;
> https://www.slideshare.net/danbri/schemaorg-and-one-hundred-years-of-search
> ).
> >
> > This has been a very backwards-looking email, and probably a bit too
> self-centered. I have been thinking about what kind of RDF/SW/etc interest
> or community group I might still be excited to be part of in another 10 or
> 20 years, and about many of our earlier contributors whose work is all but
> forgotten today. And then, about lessons for RDF in 2018?
> >
> > I think the first is that we are at our strongest when we build bridges
> between different perspectives and work roles. RDF has roots in KR, and
> strong links with academia/science, but also and equally it has roots in
> hacker/opensource culture, in the user-centric values of the library
> community, and in digital library, opensource and industry initiatives to
> integrate information and make data available through the use of simple
> standard metadata records. RDF has from the very outset been grounded in
> the great technology-meets-society debates of our time, and perhaps has had
> most traction when it was closest to efforts at the core of the Web:
> browsers and search.
> >
> > Another example that feels quite core to the RDF mission (to me). Twenty
> years ago Kjetil Kjernsmo and I were debating how best to use RDF for
> rating online misinformation (
> https://twitter.com/i/web/status/957841302070222854); this year we're
> still discussing that (e.g. via http://schema.org/ClaimReview), but we're
> also discussing data portability using shared schemas, data shapes, and
> Solid/Inrupt. Part of the problem with RDF's adoption as a technology is
> that it sometimes comes across as a bundle of burdens, not just a simple
> technology standard. Adopters have sometimes reported that they feel
> pressured by our community into jumping through various hoops - whether
> around namespaces (use lots! don't invent your own!), identifiers (always
> use them for every entity mentioned, and re-use good ones!), reasoning
> (don't break it with scruffy data), data (public, persistent, open, ...),
> HTTP headers etc. But the community/society aspect can be a strength too.
> We are working with a technology which is at its best when multiple
> over-lapping datasets are superimposed, and for that to work well, there
> has to be an element of community collaboration that goes beyond the strict
> technicalities of the W3C specification.
> >
> > In those earlier years - especially the difficult period through 1999
> through until the Semantic Web Activity was kickstarted in 2001 - the RDF
> Interest Group was the place to be for RDF stuff. We even had a few f2f
> meetups (https://www.w3.org/2001/02/rdfig-f2f/). As the standards matured
> and conferences, journals and other fora arose, we kind of stepped back and
> the list has become a lot quieter. I don't think it needs to attempt to be
> in 2018/9 what it was in the 1990s and early 2000s, but there may be
> values, themes and interests than continue, and that deserve a home. Again
> in earlier times, RDFIG/SWIG was an important focus for early (pre
> standardization even) interop testing amongst RDF query/database
> implementors. It was where we explored common schemas for things like
> calendar interop. It was where Brian McBride led much of the issue-tracking
> work that later fed into the RDFCore WG. It was where Alistair Miles and
> others took earlier RDF Thesaurus designs (
> https://www.w3.org/2001/sw/Europe/plan/workpackages/live/esw-wp-8.html)
> and reworked them into the earliest versions of SKOS. In 2018, starting
> fresh, all of these things would simply be independent W3C Community
> Groups. 15-20 years ago, we didn't have that mechanism available to us, and
> so everything was squeezed into one big (RDF/SW) "Interest Group".
> >
> > In practical terms, is there a need for a successor to RDF-DEV, RDFIG
> and SWIG? Perhaps not. W3C in 1997-1999 and the early 2000s was a different
> place. In many ways the efforts of this community trail-blazed W3C's
> subsequent opening up into public-participation groups, through blogs,
> wikis, public logged IRC chat, collaborative technology projects, and
> suchlike.
> >
> > That is in fact the reason for this week's SWIG shutdown. We have gone
> from being ahead of our time (the rest of W3C opened up years later), to
> being a kind of legacy historical anachronism, a glitch in the
> spreadsheets. The W3C Community Group mechanism, in some ways a result of
> our early experiments here with public participation, is just the modern
> way of doing this kind of thing, and we should catch up with the new
> mechanisms. It simply doesn't make sense any more to have a public
> participation Interest Group on W3C's management books in 2018. In
> addition, the new CG infrastructure, which many of us have already been
> using for related groups, offers facilities that we don't have for SWIG,
> including some IP commitments which make it easier to produce W3C documents
> together, as well as useful tools like blogs.
> >
> > There are already a ton of W3C Community Groups on all kinds of topics
> relating to the interests of folk here. You can check out the list at
> https://www.w3.org/community/groups/ if you've not looked recently. It is
> a relatively self-service mechanism, if you can get a small group of
> interested parties together who'd like to have a group, the bureacracy is
> pretty lightweight. That's what I'd like to do here with the group that
> began as RDF-DEV, and which was RDFIG and SWIG for many years.
> >
> > Ok, this mail is too long. I have filled out the proposal form for a
> RDF-DEV CG,
> >
> > "RDF-DEV, for developments relating to W3C RDF, including collaboration
> around applications, schemas, and past/present/future related standards.
> Successor to SWIG/RDFIG."
> >
> > The list creation bot says "CONGRATULATIONS! RDF-DEV is now in our list
> of proposed groups with you as the first supporter."
> >
> > See https://www.w3.org/community/groups/proposed/ for the link. I
> promise not to send mails this long if the group goes ahead.
> >
> > Regarding mailing lists, I would encourage the W3C team to leave this
> one (semantic-web@) alive, alongside public-lod@,  and then we can use a
> new rdf-dev@ list bonded to the CG for those who want to collaborate more
> closely, e.g. on W3C specs. But I think it's best left to Ralph, Ivan et al
> to make a judgement call on that, based on the responses in these threads.
> >
> > cheers,
> >
> > Dan
> >
> >
> >
> >
> >
> >
> >
>
>
>

Received on Tuesday, 16 October 2018 17:16:16 UTC