h a l f b a k e r y
I CAN HAZ CROISSANTZ?
add, search, annotate, link, view, overview, recent, by name, random
news, help, about, links, report a problem
or get an account
A browser that makes hyperlinks between data records meaningful to make. (Internet of Data)
The idea that any record in any database could have a "url"
to any other record in any other database ("Internet of
Data") may be useful in many situations, including the
creation of semantic web.
Given a good popular data browser, many would probably
care to open their databases for public
access with default
database ports, like 5432, or 27017, for data consumption,
and so, we could start using hyperlinks like:
postgresql :// www.example.com:5432/example/topics/123
The database programming languages like (PL/SQL, PSL,
would come back to relevance, allowing the flexibility to
define the added business logic and functionality, which
these days usually is provided by building an API on top of
Today, while most of the web pages on ports 80/443
are open to public, traditionally, most records in databases
are behind an authentication, and thus, people are not
inclined to make such links (they are inclined to scrape the
web rather than properly use databases).
Given an existence and wide knowledge of a good data
browser (with ease of use and powerful analytics
capabilities on local computers), it is reasonable to expect
people would care to make proper public access to them,
and create a web of data.
Some talk about where it would all go. [mylodon, Oct 23 2017]
A great big list of open databases queryable by SPARQL If the WWW is all the linked web-pages in the world, this represents a formative top-down view of all the linked-data-web as it currently stands. [zen_tom, Oct 23 2017]
A query language for linked data supporting federated queries, amongst other things. [zen_tom, Oct 23 2017]
TBL discussing linked data 10+ years ago.
[zen_tom, Oct 23 2017]
Python Library: rdflib
Full sparql support in case you want to write your own linked-data browser. [zen_tom, Oct 23 2017]
RDF/Triple-Store browser that supports federated queries - nice thing about this one is the direct visualisation view, though has a fairly hefty license tag. [zen_tom, Oct 23 2017]
||The only problem is getting people to standardize on something or to open databases to something. It's a very untrusted world out there. For instance PDF's are how most people exchange very important information, and those things are two steps from a pcx file, and you can't even palette cycle them.
||If you want to go one better, you get CSV.
||It's time to re-read "The Memorandum for members and affiliates of the intergalactic computer network."
||The difference between this and say a SQL over Telnet is .... ?
||[FlyingToaster] This would have to implement some
kind of generic query language, and ready-made
queries, and maybe visualization capabilities not seen
in command line. Communal databases like HB would
be way more interactive :)
||I was under the impression SQL is a "generic query
||There is any number of tools today that will attempt to
create analytically useful data from queries.
||SQL (at least the variants I'm familiar with), already
supports database.table syntax in queries, so that
queries can be made across databases, and there's no
particular reason why the database part of the name
cannot be a URL.
||Modern SQL also returns data in XML and can be made to
||So the principal idea here is to open the databases to
querying and to have a DNS like super table of
||How would block-chaining prevent the inclusion of fakes I
||Psst! You're describing almost exactly corresponds with
the current state-of-the-art in linked-data, as originally
suggested by Tim Berners Lee around 2006.
||The analogue for html is rdf, and the world-wide-linked-
implemented using servers running an rdf-native query
language called SPARQL.
In the public domain, there is an open, queryable version of wikipedia that's
available via federated SPARQL query, plus a host of
other life-science, language, and other domain-specific
||Where I work, we use it to keep track of enterprise
architecture and organisational components (applications,
racks, servers, facilities, staff, management,
departments, audits, regulations, risk-events, market-
events, and other stuff) in a federated set of open (or
shut, depending on the security context) RDF databases
accessed over the network by SPARQL.
||What's neat about this is that
while each individual application owner can publish an
RDF version of their data (or allow us to generate one) it's
a relatively light-touch-task.
||However, once done, it means it's finally possible to
situationally query what business people will be affected
due to a server outage in a particular building, identify
which of their clients are most likely to be affected
negatively, and send a report containing the law offices
of the most litigious of these to the legal department as a
||Previously, that query would have been impossible, but
with rdf data collected from applications, their data-
flows between one another, the it hardware estate,
organisational and managerial trees, business process
models, client lists, business activity, KYC data and legal,
and the ability to stitch all these together into a
meaningful line of dependency, it becomes possible.
||We've written a few browsers, but to date, my favourite
RDF browser tool is something called Gruff, by a company
called Franz. Though, for real-in-depth analysis,
opensource libraries (like rdflib for example) provide the
||[zen_tom], you correctly noticed it, Tim Berners-Lee
highlighted its importance many years ago. So, why,
when so many are talking about IoT, few are aware of a
bigger idea of Linked Data? Maybe it's not properly
marketed, maybe we should call it not "Linked Data",
but "IoD"? But,.. just maybe, the problem
is that we don't have that Data Browser?
||I think it's a few different things at once - and I concede
the point, a key element of that is the browser - if you
can democratise the experience, make it simpler for
people to work with linked data, the result will be more
users, more take-up of the idea.
||Having seen things like Gruff, and other subject-
predicate-object node-driven tools (lodlive etc), the
dynamically constructed graph seems to be a common
model. I've experimented with projecting these node-
graphs into 3d using (for example) minecraft as a browser
- which was fun for a bit, but the data needs to find ways
to organise itself in ways that the user can control, and
(for example) minecraft was too static for that to
||Another idea we have for our own visualisation tool is
setting different levels of aggregation, where at a given
zoom level, properties of key nodes are considered to be
parts of a 'cloud' representing a single entity, which splits
apart at differing levels of enzoomedness. The trick
seems to be tagging different property sets as having
differing levels of specificity.
||There *are* data visualisation tools, things like Tableau,
Qlikview, d3 and others. Common things you might want
do with this kind of data is explore shapes, determine
clusters, things that are similar to other things in unusual
or unexpected dimensions - which suggests network
diagrams of nodes connected by edges arranged in such a
way to demonstrate the structure. Force Directed Graphs
are one common layout device that might be a data-
browser default setting, but I think such a browser should
auto-detect common network shapes such as trees,
directed-flow-networks, clustered arrangements etc and
be able to automatically pick a layout that best fits the
||If I'm looking at a directed flow, ideally, I want all the
sources on the left, and all the final sinks on the right, (or you might prefer up-
down, or right-left
orientations, or a clockwise/anti-clockwise view, whatever)
with some intelligent placement of intermediate nodes
such that the strongest flows are evident, Sankey style.
This is one such layout option that could be inferred from
a dataset and offered to a casual user.
||The point being - in our very small department at work - one line we're
exploring is just this idea - common, generic methods for visualising linked-
data queries in the browser that aren't tightly-coupled to the data being
queried (otherwise, they'd be called 'reports' )
||So I'm very excited by this idea, as it's one of the first times I've seen this
type of thing talked about outside of my department at work. There are I
think quite a few isolated pockets of people working on this problem
though, and already people like the BBC, the UK and US governments and
many others have all invested quite heavily in RDF/Linked-Data as a back-
end protocol for knowledge-management. There are front-ends out there
for visualisation, and things like Gruff do a decent job, but there's a long
way to go.
||//employ a lot of budget creating hand-waving// hand-waving is one of my
specialities, when this coincides with budget-creation scenarios, then that's a