Re: Queries on Single Sourcing

["Mark Baker" <listsub -at- analecta -dot- com>]

Lyndsey Amott wrote:

> This guy had implemented in his group a system such as you describe. The
> idea was that you could re-use any chunk of text, assuming that you could
> find it in the huge repository. In fact, writers were required to look for
> it in the repository, whether it was there or not. The fact that, when and
> if it was found, it had to be tweaked, and the fact that I have rarely
> re-used an untweaked chunk of text in my entire career, and the fact that
if
> he got hit by a bus no one would be able to figure the system out seemed
to
> be lost on him.

Systems like the one you describe are the result of a failure to understand
the difference between a jigsaw puzzle and a lego set.

A jigsaw puzzle takes a picture and cuts it up into little pieces. However,
the only picture you can make is the original one, and it takes a lot of
work to put it back together again. You can't reliably use the pieces to
create another picture. Nor can you mix the pieces from several jigsaw
puzzle pieces together and hope to create a bunch of new pictures by mixing
and matching the pieces from different sets.

A lego block, on the other hand, is not part of any one model. It is
specifically designed to be usable in a variety of different situations. It
has a well defined size, shape and interface so that it can be combined
easily with other types of lego blocks. Lego blocks can be used to build a
wide variety of different structures and vehicles because they were designed
specifically to be used like this.

If you want to build a repository of reusable / single source information
components, you have to design them like lego blocks, not like jigsaw puzzle
pieces.

And it is important to understand that the difference between a lego block
and a jigsaw puzzle pieces is not metadata. You can add metadata to a jigsaw
puzzle pieces (just write a set of co-ordinates on the back) and the
metadata will make it much easier to assemble the original picture. But no
amount of metadata is going to make it possible to mix up several jigsaw
puzzle pieces and use them to create new pictures.

You don't make lego blocks by buying a collection of plastic toys and
hitting them with a hammer until you have broken them into small enough
pieces. You also do not make lego blocks by buying a bunch of plastic toys
and cutting them up with an Xacto knife and making a pile of windows, a pile
of chimneys, a pile of wheels, and a pile of left arms. Categorizing the
pieces into those piles is applying metadata to them. It leaves you with a
well organized collection of discrete components that still do not fit
together properly to make new toys.

You make Lego blocks by designing them as lego blocks from the start and
building them to strict Lego block standards. If you have a small collection
of Lego, you can work with them quite easily by just spreading them out on
the carpet and selecting the block you want. If you have a large collection
it will be useful to sort different kinds into different boxes (that is,
apply metadata to them).

Metadata is useful for organizing your well designed information components,
but no amount of metadata will turn a document fragment into a reusable
information component. With a small collection of information components you
can manage with little or no metadata. With a larger collection, metadata
becomes more important to help locate the components you need. On the other
hand, you can assemble a repository of document fragments and no matter how
much metadata you attach to those fragments, you still won't get any reuse
out of the system.

Mark Baker
Analecta Communications