With the ever growing use of XML as a data representation format, we see an increasing need for robust, high performance XML database systems. While most of the recent work focuses on efficient XML query processing, XML databases also need to support efficient updates. To speed up query processing, various labeling schemes have been proposed. However, the vast majority of these schemes have poor update performance. In this paper, we introduce a dynamic labeling structure for XML data: L-Tree and its order-preserving labeling scheme with O(log n) amortized update cost and O(log n) bits per label. L-Tree has good performance on updates without compromising the performance of query processing. We present the update algorithm for L-Tree and analyze its complexity.