Naming is a Problem
By Cris Fitch

In the following paper, I discuss why I think Naming is so important
in computer programming, and how one might tackle the problem of
creating a sophisticated program to generate names for use in
programming, creative writing, or perhaps most importantly, for
use on the Web.

Why Naming is Important

It is my contention that Naming is an important topic to cover in
teaching programming.  A name can serve to direct the conception of
the entity you're constructing.  A generic name, such as foo, can be
used to hold the place of a more appropriate name that may come along
later.  A very descriptive name, however, may serve to distract both
the creator and the maintainer of the code, since the description may
have gone stale.  A creative and unique name may enliven an entity,
although care should be taken to use these sparingly.

A popular country music song of my youth told the story of a "Boy
Named Sue".  With a girl's name, the boy grows tough, having to
continually respond to the taunts of others.  As a he becomes a man,
he tracks down the man who gave him this name, his father.  After
finding this man, and beating him in a vicious fight, he demands to
know why a man would name his son "Sue".  The father explains that, as
he knew he would not be able to look after his son, the name
would force him to be tough and thus be able to survive the world's
harshness.

In the process of designing and coding computer software, the creation
of names is a frequent activity.  Once created, a name tends to remain
an integral part of the program throughout that code's lifetime.
For code to be self-documenting and easy to maintain, these names
end up being key.

One can name something in a variety of different ways.  Normally you
want to choose something that is both unique and appropriate.  Perhaps
even descriptive.  As in the story, a bad name can cause one much
grief.  On the other hand, it can lay the foundation for meaningful
reference, perhaps the ultimate purpose of a name.

The Power of a Name

When one knows the name of something or someone, it gives them
a certain power over that something or someone.  It is the power
of reference.  With a single token, a name, one can grab hold
and manipulate the essence of an entity, without having to describe
in detail the parts therein.

Most people know the story of Rumplestilskin.  In it, a dwarf saves a
maiden from certain death when she is unable to weave straw into gold,
as demanded of her by the king.  In return for a bit of magic, the
dwarf gets the girl to agree to give him her first born child a year
hence.  The king marries the girl, and happily she has a beautiful
child.  When the dwarf shows up to collect his payment, the girl
refuses.  He gives her the option of guessing his name, instead.  This
she learns by secretly overhearing him.  When, on the next day, she
successfully provides his name, his power is broken and he disappears,
never to be seen from again.

In this era of the World Wide Web, the story becomes highly relevant
as people attempt to find things in an ocean of knowledge.  Search
engines for the Web index tens of millions of pages, and searches must
often hinge on key strings.  With a name like "John", how likely are
you to be found?

In considering the construction of a program whose goal is appropriate
naming of entities, there are challenges in how to measure
appropriateness and how to generate potential candidates.  Both of
these are highly dependent on the reference material which is
consulted.

Where do Names Come From?

Naming needs to have some references upon which it is based.  A
lexicon, a collection of common-use words, is an important foundation
herein.  Each word within the lexicon may or may not have a unique
string which indicates it.  Associated with a given string, as in
WordNet, is one or several concepts.  These have a part of speech, a
place in a hierarchy of concepts, and other various properties.
Extending our reach outside the lexicon, there are a world of
associations that common-use words and names have.  When used in a new
fashion, as part of a program for example, a name may come to acquire
new associations.  Yet those old associations remain and can either
help or hinder the new role you desire a name to play.

A subtle distinction should be made between single and multi-word
names.  Whereas a single word name can either appear or not appear in
the lexicon, the multi-word name has a constructive aspect to it.
With a finite lexicon, multi-word constructions can be infinite.

For single word names not found in the lexicon, a different
associative mechanism kicks in.  Pronunciation and the similarity of
appearance to items which are in the lexicon has an impact.  Gwity may
be a new word.  Yet the ending (morphology) and its similarity to
Gritty or Witty may be evoked.  In other words, no name is completely
without association.  Yet the reason behind a completely new name (not
in the lexicon) is the emphasize uniqueness and to build up a totally
separate identity for your construct.

Within a standard American lexicon, there are words that tend to be
names, and those that tend to be common use English words.  Names
themselves break up into several categories, as strings which tend to
be appropriate for places, people, companies, etc.  Some of the
shortest names are acronyms, such as IBM or GM, although these can
often lose their uniqueness, such as NRA, PLA or PRC.

Even if not present in a dictionary or encyclopedia, a name can be
present in the popular literature or culture.  Consider the use of
names from the Simpsons or Tolkien.  With the use of such a name,
associations may be helpful, either for marketing, memory, or
extension, or they may be cryptic, obscure or irrelevant.

Obviously an English parsing lexicon is a good place to start, since
it serves to map many commonly used strings.  It might also make sense
to have a large set of commonly used names, and some sense of where
they have been applied.  Associations between strings gets to be a
difficult thing to tackle because of the immensity of the knowledge
involved.  However, the existence of the Web serves to further this
cause much more than we might have been able to consider even a few
years ago.

A Naming Mechanism

The first thing that comes to mind when asked to write a program which
names things is to build a random string generator.  I have seen
Cognitive Science experiments which display random English-like words
for measuring recognition speeds.  These generators used statistical
frequencies to adjust the kinds of output.  For example, the letter
'e' appears more frequently at the end of a word than at the
beginning.

But a name isn't simple a new, unique string.  I call my cat 'Siberia',
which uniquely identifies her from other cats.  But it also has
additional meaning and implications.  When asked why I had named
her that, I tell them that her fur has the same color as the bear
on a PBS documentary about Siberia which was showing at the time
she adopted me.  And when I took her to the Vet, and they wrote her
name down, they wrote it as 'Cyberia', with a whole separate set
of implications.

Returning to the random selection approach, we wonder whether it
might be possible to just pick a random item in the lexicon.
Sometimes this might be a good thing to do, and in the case of
my cat, it was nearly the case.

Names and Meaning

Is a name random, though?  Once established, associations develop
and justifications are put in place.  A story may be formed to
explain the choice of a given name, in reaction to people's need
to have an explanation.

Yet for the selection of names for software entities, this approach
alone could cause chaos.  It denies the pre-existence of meaning
for items within the lexicon.  If the cat came from a sister living
in Cincinatti, one wouldn't want to call the cat 'Chicago' or
'Buffalo'.  True, the set of potential names is large.  But
appropriateness has some relation to the pre-existing circumstances
of this thing without a name.

It is important to note which items have names.  One doesn't uniquely
name the speakers hooked up to your stereo or the loaf of bread you
bought at the store today.  Why is that?  In the days of yore,
peasants were often named simply by what they did.  Baker, Smith, and
many others (my own included) were job descriptions as much as names.

In programming this is true of iteration variables, for example (for
i=0 to 10).  It doesn't deserve a globally unique name.  It is perhaps
best that the name of the item is something so simple, so common.  And
by having that simple name, when you encounter it, you already know so
much about its place in the world.

I've found in my own code that there is a time and a place for a unique
name.  It serves to balance the blandness of so many generically
named items.  Especially when the meaning of the entity has yet to
be established and yet will cover a large section of new territory.
The most obvious case is the name of the overall program itself.
Many of the utilities I write have unique overall names, because
that way they can develop their own personalities, independent of
some foolish initial perconception.

The major alternative to a unique name is a functional description.
Would it have been that strange if Kodak had instead been named
General Camera?  Instead of calling my new utility program "Waxito"
I might name it "FilmConvert".  It takes less time and effort to
create a unique name, and it doesn't become obsolete when the program
evolves away from a narrow original intent.  Yet to the uninformed,
it has little relevant meaning and requires that a new entry be
created and maintained in the lexicon.  A functionally descriptive
name is "self-documenting".