Algorithm note

I saw a post on one of my coding boards talking about a solution for an 'interview-style' problem.

The issue was to find the longest common substring.

What this means is: Given the word "abracadabra" you are looking for the longest repeated set of characters.

The longest repeated sub-string is "abra" which appears at positions 0 and 7 (counting from 0).

The solution the person presented involved putting the strings from smallest to largest into a dictionary (hash table) with counts of iterations.
a-4
b-2
ab-2

And then walking the tree to find the largest result. This solution seems odd to me.

My solution is to make a dynamic array that can be queried for contents. Further, doing it from longest to shortest:

Edit gwenix points out a string can't be repeated until it's no more than half the length. So start from longest string (len /2) rounded down.
End Edit

Solution:
Get longest string: abracadabra
Is string in array? YES? got answer and bail; No --> add to array
*Reduce look string by 1 character
Move to start of target string:
**Get look string: abracadabr
Is string in array? YES? got answer and bail; No --> add to array
Is look string at end of target string? Yes --> loop back to *; No --> move forward one character and loop back to **

This is massively not optimised for time or processing. But it's a first stab.
Also, just pseudo code... loop counters would be my next pass.

Comments? Improvements?