Hi Manu,
On 06/13/2013 03:50 PM, Manu Sporny wrote:
bcc: RDF WG
*JSON-LD Chair hat on*
David Booth wrote:
In normal usage within the Semantic Web community, does the term
"Linked Data" imply the use of RDF?
David,
I really wish you would have passed this poll by both the JSON-LD group
and the RDF WG before sending it out so that we could have settled on
the set of right questions to ask.
And spent even *more* time arguing over the wording of the poll? I
thought that would be a bad idea.
But other polls can certainly be developed, and I would encourage others
to do so if they wish. If we get enough answers from them, the results
could be very illuminating in understanding the different perspectives
in different communities.
As others have pointed out, you're
phrasing the question in such a way as to ignore the point of
contention.
Apparently we have vastly different ideas of what that point of
contention is.
This is only going to muddy the already opaque waters.
I do not see how quantitative data on the meaning of a term could
possibly muddy the waters.
Additionally, I don't see what marginalizing other Linked Data
technologies (that are "not RDF") is going to accomplish.
I have to object to that statement, because the phrase 'other Linked
Data technologies (that are "not RDF")' already implies that Linked Data
is *not* necessarily based on RDF, whereas that is the question that
this debate is trying to resolve!
To the rest of the members of these mailing lists,
A number of us have been having an off-list exchange over why the text
outlining Linked Data is what it is in the JSON-LD spec. I'm going to
share parts of that exchange (only the bits that I wrote, names
redacted) in order to shine some light on why David Booth is trying to
get feedback from this community.
------------------
[Assertion that Linked Data implies the use of RDF]
Manu Sporny wrote:
I know that you think it [Linked Data] has a well-established meaning
in "the community". I disagree, not because I necessarily care that
deeply about it, but because when we asked "the community" for the
definition, we received a myriad of different answers from people
that have been core to the process over the last decade. This led to
a huge amount of work for the group because of exactly the type of
exchange that is happening between [XXX] and [YYY]. [The definition
of Linked Data] may seem clear to each of you, but I can assure you
that it's not at all clear to the rest of us.
Hopefully this poll will shed some quantitative light on what this term
means **within the Semantic Web community.** It will not answer the
question of what the term means to those outside of that community who
have little or no knowledge of the term's origin.
Both of you are arguing over what each of you think was in TimBL's
head at the time. We can't base anything off of such a line of
argumentation.
The argument has been approached from multiple angles: assertions about
what the term currently means, evidence about what it currently means,
evidence about what TimBL coined it to mean, and contextual
recollections about what was going on at the time. the assertions about
what the term currently means are likely to go nowhere without
quantifiable evidence. that's why I made that poll. other less direct
evidence of what the term currently means -- looking at the top 10 hits
of the three most obvious Google searches of the term's meaning -- did
not seem to have any affect on those who had already made up their
minds. (perhaps they never looked at them?)
Honestly, the problem seems to be that there are some loud voices that
are not even *looking* at the evidence, regardless of how convincing
that evidence may be.
case in point: I have been repeatedly been accused of equating Linked
Data with RDF, which is patently untrue. Regardless of how many times I
have pointed out that I have *never* equated Linked Data with RDF --
that Linked Data is RDF but RDF is not necessarily Linked Data -- it
seems to have no effect, which is quite frustrating.
another case in point: I have multiple times seen the assertion that
TimBL's original Linked Data document did not mention RDF, in spite of
the fact that anyone who takes the time to actually *read* that document
can plainly see that it *explicitly* mentions RDF:
http://web.archive.org/web/20061115043657/http://www.w3.org/DesignIssues/LinkedData.html
this certainly makes it appear as though some voices in this argument
would rather stick to their opinions than look at the actual evidence,
which is rather discouraging in a technical forum.
We debated the normative definition of Linked Data for months during
the development of JSON-LD. We even came up with normative spec text
so we could define what Linked Data is once and for all. We were
then asked to remove the definition from the spec by the RDF WG
because "it may be used to exclude certain technologies as Linked
Data and further fracture the Linked Data movement." Damned if you
do, damned if you don't.
You can read the logs here, if you'd like:
http://json-ld.org/minutes/2011-07-04/
http://json-ld.org/minutes/2011-07-26/
You can see the start of the normative definition of Linked Data
here:
http://json-ld.org/requirements/ED/20110703/#linked-data
We did this because there was, and continues to be, no normative
definition of Linked Data.
You are going to get a fantastic amount of push-back if you claim
anything to the contrary on the call. I'd re-think your approach if
I were you and base a line of argumentation on why "RDF" should be
mentioned on first principles rather than relying on something TimBL
wrote over a decade ago.
He wrote that Linked Data document in 2006 -- not over a decade ago. I
am not basing my argument on that document. I am basing it on the
established meaning of the term within the Semantic Web community and on
the importance of the term to the Semantic Web community. That document
only provide evidence of the term's original meaning within the Semantic
Web community. It does not provide quantitative evidence of the term's
current meaning within the Semantic Web community.
[Assertion that the JSON-LD group is deliberately misleading the
public by not asserting that Linked Data implies the use of RDF]
Manu Sporny wrote:
I can't speak for the others, but I'm not convinced that we should
change it at this point because your line of argumentation has been
unconvincing. All that has happened to date is that both you and
[YYY] have provided evidence that there exist versions of TimBL's
document that do include RDF and don't include RDF when the concept
of Linked Data was introduced.
Again, that is factually incorrect. I implore you to *please* read that
document:
http://web.archive.org/web/20061115043657/http://www.w3.org/DesignIssues/LinkedData.html
Don't take my word for it -- just read it. It very clearly *does* say
RDF. It does not say RDF in *every* sentence, because RDF was the
context. But notice the very first sentence. And notice the full
paragraph that explains rule three of "The four rules":
[[
The third rule, that one should serve information on the web against a
URI, is, in 2006, well followed for most ontologies, but, for some
reason, not for some major datasets. One can, in general, look up the
properties and classes one finds in data, and get information from the
RDF, RDFS, and OWL ontologies including the relationships between the
terms in the ontology.
]]
Notice that it says "*the* RDF, RDFS, and OWL ontologies.
Again, it is quite frustrating to repeatedly read verifiably false
statements about that document, when the document is right there for
anyone to read.
Additionally, you haven't countered these pieces of evidence at all:
http://www.w3.org/DesignIssues/diagrams/history/proposal-fig1.gif
Wow. I am baffled at why you think that needs to be "countered". That
is an illustration from TimBL's original proposal for the web. It
pre-dated TimBL's coining of the term "Linked Data" by over a decade! I
have no idea why anyone would think it is relevant to the definition of
Linked Data.
http://www.w3.org/2005/Talks/1110-iswc-tbl/#(4)
http://www.w3.org/2005/Talks/1110-iswc-tbl/#(7)
Those slides also pre-dated TimBL's coining of the term by nearly a
year. Again, I have no idea why anyone would think they are relevant to
the definition of Linked Data.
http://dig.csail.mit.edu/2007/Talks/0511-tab-tbl/#(10)
That presentation is about Tabulator, which is an *RDF* browser! The
fact that slide 10 did not mention RDF should not be surprising, because
the *entire* slide set is about RDF! (Take a look!)
Similarly, [YYY] hasn't countered yours. It's impossible to do so.
Neither one of you are going to be successful in convincing the other
about what was going on in TimBL's head. So, we're just stuck here
watching [XXX] and [YYY]'s conversation devolve to the following on a
public mailing list:
Yeah-huh! Nuh-uh! Yeah-huh!!! Nuh-uh!!!
You need some new, convincing evidence.
That is exactly what I am trying to obtain by putting out this poll.
Additionally, even if that
evidence is TimBL himself stating that Linked Data is based on RDF,
you should prepare yourself for people to disagree vehemently with
that assertion as there are research papers that stretch all the way
back into the late 1970s covering things that pass the duck test as
Linked Data:
1976 - Peter Chen's thesis on "Linked Data"
http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.123.1085&rep=rep1&type=pdf
What???? How could a 1976 paper that never uses the term "Linked Data"
possibly have any bearing at all on the current meaning of the term
within the Semantic Web community?
This argument is not about whether there exists data that is linked
without using RDF, nor is it about whether such concepts existed before
TimBL coined the term "Linked Data". This argument is about what this
term now means -- as a term of art -- within the Semantic Web community.
[[YYY] has misled the community into thinking that Linked Data
does not necessarily imply RDF.]
Manu Sporny wrote:
What convinced me were facts and a sound line of reasoning. You are
assuming that the people that work on JSON-LD are easily swayed. You
have used facts, but have not provided a sound line of reasoning for
why we should state that Linked Data is RDF or is based on RDF.
I don't understand what you mean. If facts are not enough, what kind of
line of reasoning do you want?
The very next e-mail I saw was the one calling for a poll on this matter.
The point of contention isn't about what the Semantic Web community
thinks about the definition of Linked Data.
Apparently we have very different views about what is the point of
contention here, since to my mind that is *exactly* the point of
contention.
The point of contention is
if a large majority (90%+) thinks that Linked Data requires/implies the
use of RDF.
Majority of what population? It would be pointless to ask a general
population, since we are talking about a particular term of art. if you
dilute the population with those who do not know what the term means, or
who *think* that they know what it means, based on its generic
descriptive meaning rather than its special meaning as a term of art,
then you get a distorted and misleading answer.
In order to discover the answer to that question, we would need an array
of questions posed to the general Web community that ask the following:
The general Web community??? That would be like asking the general
population to define the medical term "acute care". (Most of the
population thinks that term refers to treatment for a severe medical
condition. But it doesn't. It means short term medical care.)
Asking the general Web community will shed no light whatsoever on the
special meaning of "Linked Data" as a term of art in the Semantic Web
community.
1. Does the use of Linked Data imply the use of RDF?
"Of course not. My database tables are linked, and they don't use RDF.
What is RDF anyway? Isn't that like an electronic barcode or something?"
2. Does the use of Linked Data require the use of RDF?
3. Do you need to utilize many of the RDF Concepts to publish Linked
Data, or are simple key-value pair objects that use URLs to refer to one
another enough? That is, if JSON-LD didn't map to RDF at all, would it
still be considered Linked Data?
4. Must the definition of Linked Data make it explicit that Linked Data
is based on RDF?
5. Is there an EAV model that is not the RDF model but could be used to
express Linked Data?
6. Can vanilla JSON express Linked Data?
We can't address this question without being nuanced, because the
nuances matter. We can't focus the questions on just the Semantic Web
community. We need to cast a wider net and bring in Web developers with
no real background in formal logic and semantics.
We already went through this process before and whittled the definition
of Linked Data down to something that achieved consensus, and stood
untouched for a long time:
http://www.w3.org/TR/json-ld/#introduction
Going around the group, re-raising the question while submitting no new
evidence, especially after all of the evidence was laid out plainly, is
bad form.
There was no attempt to go around the group. This was an attempt to
efficiently add some objective, quantitative evidence to the debate --
pure and simple -- to reduce the subjective back-and-forth arguments.
That said, if you would like to work with the JSON-LD and RDF groups to
formulate a questionnaire that meets the needs of both sides of this
debate, I urge you to do that (and not circulate polls that only have
one correct answer).
We could, and I would be happy to help, but it sounds like we so far
pretty far apart about what questions should be, so it may take some
work to figure that out.
BTW, lest anyone gets the wrong impression of where this discussion is
going, I would like to quote from another private message in the
exchange that Manu mentioned. This one was from me:
[[
I think everyone's intentions are in the right place -- including
Kingsley's -- but in some cases I think zeal overstepped accuracy.
Fortunately, I think we are on a good path to correct those oversteps
while still being developer-friendly. I personally appreciate the
group's willingness to work directly with members of the community to
resolve these issues fairly.
]]
That still reflects my overall view.
Thanks,
David
