% BE SURE TO EDIT design.ptex
\documentclass{article}

\usepackage{rcs}
\RCS $Date: 2000/03/04 15:45:59 $
\RCS $Revision: 1.1 $
\date{Rev.\RCSRevision~~\RCSDate}

\title{Dimension-based thinking for ZigZag}
\author{Tuomas J. Lukka}
\begin{document}
\maketitle

\def\zz{ZigZag}

\section{Introduction}

When designing the code to implement a \zz\ space, the 
place most people would start at would be the cell object, which would
have a set of pointers to other cells, named with dimension names.
However, this is not the only way to structure the internals - indeed,
it is not necessarily a good way at all.

\section{Dimension-based view}

With cells, you have to always take steps to enforce the two-directional
connection consistency rule.
The dimension-based view starts from the realization that if we consider
each {\em dimension} an entity by itself, this gets much simpler: 
simply put, each dimension is a one-to-one mapping from some cells of the
space to some other cells. 

\section{Advantages}

The dimension-based view has several advantages:
\begin{itemize}
\item {\bf Specialization}
	It is much easier to define a single dimension object to 
	do something different than the others, e.g. a dimension where
	getting the {\it n}th cell is fast but inserting and deleting cells
	is allowed to be slow.
\item {\bf Simplicity}
	Since one object is in charge of the both ends of a connection,
	it is easier to avoid mistakes. Coding transient cell connections
	is easy as well, as that object can keep tabs on the transient
	connections separate from the stable ones.



% .PS
% box "Clang" ht 0.25
% .PE
% $$\box\graph$$