On 6/24/13 9:12 AM, Antoine Isaac wrote:
On 6/24/13 2:44 PM, Kingsley Idehen wrote:
On 6/24/13 6:23 AM, Antoine Isaac wrote:
Hi Dominic,
I agree with the relevance of the effort, and wouldn't argue against
centralizing. Not everyone will have the resource to search in a decentralized
fashion...
What worries me a bit is how to learn lessons for the past. As you (or someone
else) has pointed, there have been previous attempts in the past.
For example http://www.w3.org/2001/sw/sweo/public/UseCases/
I don't find the cases there super-technical. And is it really from the past?
Looking closer, it seems still open for contribution:
http://www.w3.org/2001/sw/sweo/public/UseCases/submit.html
Actually I have submitted a case there way after the SWEO group was closed:
http://www.w3.org/2001/sw/sweo/public/UseCases/Europeana/
Now why do these things seem obsolete to newcomers?
Just giving some account on what I've been involved in ...
[Note: I'm sorry if sometimes it's going to read a bit as a rant. It's not
intended, just trying honestly to reflect the situation ;-) It's also not
purely about your case/requirement situation, but I believe the issues are very
similar!]
[Perspective from the case providers]
It's hard to know where to contribute. Existing don't often come in the places
where case owners are, or it's hard to tell whether they're still open. And
there's always a fresher initiative (like the one you're trying to launch)
which seems a good place.
In fact I have actually created some updated description of the Europeana case
http://lodlam.net/2013/06/18/what-is-europeana-doing-with-sw-and-lod/
But because the LODLAM summit was a more actual forum for me recently, I've
posted it there. And failed thinking of updating the SWEO list, mea maxima
culpa.
[Perspective from the case gatherers] I have actually be involved as
'initiator' of a couple of listing.
1. SKOS datasets (which are a kind of 'case for SKOS')
We started with a web page:
http://www.w3.org/2004/02/skos/data
but as the list was difficult to maintain we soon created a community-writable
wiki:
http://www.w3.org/2001/sw/wiki/SKOS/Datasets
As it seemed not modern enough, we've then encouraged people to use the same
DataHub platform as the LOD cloud:
http://datahub.io/dataset?q=format-skos
But both are not very active. And they contain a lot of dead links...
2. Library-related datasets:
http://datahub.io/dataset?groups=lld
That list, started by the Library Linked Data W3C incubator, went alright as
long as the group was running. Now I think the rate of new datasets is really
small, even though I *know* there are many new ones.
Both as SKOS community manager and former LLD co-chair, I've tried to actively
mail people to create descriptions of their stuff. But it requires time. Most
often they assume *you* would do it!
And after a while, the supporters of such effort just have other things to do
and can't afford very high level of commitment.
What should we do if we want to build on existing lists and not re-invent the
wheel every six months or so?
Or is it worth sending a regular (monthly?) reminder to lists like public-lod,
reminding everyone that these lists are available and open for contributions?
Create a list of lists, as Wikipedia does?
Best,
Antoine
Antoine,
As you've indicated, there have been many attempts at this over the years and
they never take-off or meet their goals etc.. The problem is that a different
approach is required. Basically, in this scenario lies a simple Linked Data
publication usecase i.e., a problem that Linked Data addresses.
The steps:
1. use a Linked Data document to describe you product, service, platform,
usecase
2. publish the document
3. make people aware of the document.
Crawlers will find your document. The content of the document will show up in
search results.
The trouble is that confusion around Linked Data makes 1-3 harder than it needs
to be. Then add RDF misconceptions to the mix, and it gets harder e.g., that
you must have generally approved vocabulary before you get going, when in fact
you don't.
People need to understand that "scribbling" is a natural Web pattern i.e.,
rough cuts are okay since improvements will be continuous.
Kingsley
Two practical objection to this otherwise interesting approach.
1. For that kind of survey, as for the rest, people want trust. it will have to
be curated (I mean, besides people just putting little bits of
uncontrolled/outdated data out there), or it will fly only when thee
distributed descriptions are harvested and accessible through something like
Google/schema.org.
Btw people also want visibility. You don't say anything about step 3...
You can sign documents. You can even sign claims. Even better, claims can be
endorsed by others. These a issues naturally handled by Linked Data.
Verifiable Identity and Trust are areas where Linked Data shines.
2. It needs to be simple, as non-technical as possible. Step 1 is already too
much. Consider LD consumers, who don't publish any LD, why would you ask them
to publish an LD document?
"simple" is subjective. There are many routes to the same destination here. For
instance, some will happily craft Turtle by hand, others may do so using other concrete
syntaxes. Of course, some would prefer an HTML5 form based interface too. The key is to
be dexterous enough to handle profile variety.
Actually even in organization that publish LD having step 2 and 3 will take
some effort. Not much, I agree, but it won't be part of the core business, and
it will still need effort.
Consider the need to have an (i) updated version; an (ii) interoperable.
It can be done, the problem has been that the approaches to date don't work and
will never work. Thus, we need to try something different, one that's also a
Linked Data dog-food exercise too.
Taking a concrete example: me (again, sorry!). A while ago I've made a
description of data.europeana.eu as a voiD file. Nice, but now I hear that
there's DCAT around and I should read the doc and update my file. Oh, and my
dataset has been also updated.
There are even notification protocols that mesh nicely with Linked Data. Our problem is
that there is too much fragmentation. Suggestions to tackle these issues via dog-food and
"just do it!" patterns ultimately get lost is bizarre arguments rife with
contradiction. If Linked Data is what is claims to be, then we can address these issues
(collaboratively) via dog-food patterns.
And I've got no idea who will consume this updated file and whether it will
happen one day...
The same thing applies to any content you put on the Web, you ultimately need
an incentive to keep it up to date. The same thing applies to consumers too,
they need an incentive to want to track etc..
And I've got a hell of other more urgent things to do. So anything that won't
be populated by a trivial adaptation of the blog post, which I've already
written, will have to wait for a while.
Tweets and posts to other social media are effective mechanisms for discussion
about data that aid discovery and curation. They can also be powerful incentive
vectors.
Links:
1. http://slidesha.re/Ys79Jn -- ontology life cycle presentation I gave to some
ontologies earlier on this year (note: the presentation includes live links
too).
