deletions | additions       diff --git a/bits3.tex b/bits3.tex  index 5503a5f..fdbda1b 100644  --- a/bits3.tex  +++ b/bits3.tex 
...
\end{center}  We notice the following facts:  \begin{itemize}  \item performing \begin{enumerate}  \item\label{fact-a} After having performed  the above algorithm (seven times) seven times,  we get again the initial vector $(1,0,1)$. This is a general behaviour of a LFSR, using any polynomial.  \item \item\label{fact-b} While  performing the above algorithm, we find all the strings of three bits with at least a seven  nonzero entry. $3$-bit strings.  %only the numbers $1,2,3,4,5,6,7$ can be written using three bits not all zero and, performing the algorithm, we found all these numbers.   This \end{enumerate}  %  Fact (\ref{fact-a})  is not true in general. The feedback polynomials for which this a  property holds that  is shared by many LFSR's. It is enough to use  any polynomial $f\in F_2[x]$ of the form $f=x^n+\ldots+1$ to construct an LFSR  producing $n$-bit vectors and such that any initial state is got again after  a finite number of iterations.\\  Fact (\ref{fatc-b}) is more difficult to generalize. If we want an LFSR that can start from any initial nonzero vector and obtain all other nonzero vectors, then we must use very special polynomials,  called \emph{primitive}.  \end{itemize} \emph{primitive} polynomials.  In other words  \begin{Definition}\label{Primitive}