IEEE Consumer Electronics Magazine - October 2015 - 127

Bits Versus Electrons

The Internet Is About
Relationships
By Bob Frankston

M

y last column was about
application programming
interfaces (APIs). APIs represent a programmatic view of
systems. Devices function in a context
and in relationship to other devices, but
they do not function in isolation.
The Internet is about facilitating
these relationships as opposed to traditional networking, which is all about
"the between." To realize the benefits
of the Internet approach we need to
focus on the relationships.
A "wiring diagram" approach defining relationships between devices is
complementary to the API approach and
gives us a way to visualize relationships.
I put the term "wiring diagram" in
quotes so we do not confuse it with
physical circuits. With a wiring diagram,
we can specify the result rather than
how it is to be accomplished. A webpage is a powerful example. We link to
images and other pages without having
to be concerned about the mechanics of
fetching and displaying the page.
This ability to focus on the end points
of the relationships and not the complexities of what is between those end points
is the powerful "magic" that has made
today's Internet so transformative. Of
course, this is not sorcery but rather
defining architecture of the Internet.

RELATIONAL DATABASES
Relational databases show the power
of this approach. Early databases
Digital Object Identifier 10.1109/MCE.2015.2463413
Date of publication: 29 October 2015

were carefully laid out to optimize
performance on disks to the point that
some databases contained actual disk
I/O instructions in the database. The
database designer carefully specified
a path through the disk drives to
reach the data.
Relational databases were very different. The tables of data would be related
simply by having common identifiers-
index keys. Typically, these were just
arbitrary numbers but they could also be
names or any common identifiers. This
approach not only decoupled the data
from the accidental properties of the
hardware, but it also flattened the database to simple tables whose relationships were described by schemas.
This allowed databases to be restructured without having to change data values and it also allowed schematic
evaluation. A series of SQL queries
could be combined symbolically without
actually accessing the data, and the
resulting query would not need to access
all of the original data. One innovation
in databases is the use of a globally
unique identifier (GUID), which is
essentially a long random number that
allows disconnected databases to generate their own keys without having a central source of identifiers.
The lessons of relational databases
inform our understanding of the Internet
and the importance of understanding the
use of relationships as an architectural
mechanism. I use the term "mechanism"
to avoid referring to the relationships as
a layer on top of a file system or other
implementation. Rather, the file system

is used as a resource. Conversely today
we have NOSQL databases, which are
often defined as "not only SQL." They
can use SQL itself as a mechanism.

DESCRIPTIONS AND DIAGRAMS
The web's success owes much to similar principles. Web pages (especially in
the days before JavaScript) describe the
relationships between pages-the
links-without specifying how they are
to be implemented. The very term
"World Wide Web" builds on this idea.
The pages themselves are descriptions of how the information should be
presented without specifying how that is
to be accomplished. This is the way we
should think about the Internet as a
whole-relationships between end
points be they devices, webpages or
people. Of course, there are no real
wires tying these end points together.
The relationships exist abstractly or, to
put it colloquially, in our imagination.
When the time comes to actually follow
the link, we may find out it is not valid.
Not only is that OK, it is essentially
to the distributed nature of the web. We
learn to handle such failures gracefully.
Sure a 404 is annoying but far better
than crashing your computer.
Too bad the domain naming system
(DNS), as with the Heinz example
given in the "A Circuitous Path" subsection does not allow for gracefully
retiring names.

CONNECTING THINGS!
Today we are increasingly using the
Internet to connect things or, perhaps,

october 2015

^

IEEE ConsumEr ElECtronICs magazInE

127
Table of Contents for the Digital Edition of IEEE Consumer Electronics Magazine - October 2015
