- * except for a quirk: updates to the queue must be delayed by one LZMS item,
- * except for the removal of a repeat match. As a result, 4 entries are
- * actually needed in the queue, even though it is only possible to decode
- * references to the first 3 at any given time. The queue must be initialized
- * to the offsets {1, 2, 3, 4}.
+ * except for a quirk: inserting anything to the front of the queue must be
+ * delayed by one LZMS item. The reason for this is presumably that there is
+ * almost no reason to code the same match offset twice in a row, since you
+ * might as well have coded a longer match at that offset. For this same
+ * reason, it also is a requirement that when an offset in the queue is used,
+ * that offset is removed from the queue immediately (and made pending for
+ * front-insertion after the following decoded item), and everything to the
+ * right is shifted left one queue slot. This creates a need for an "overflow"
+ * fourth entry in the queue, even though it is only possible to decode
+ * references to the first 3 entries at any given time. The queue must be
+ * initialized to the offsets {1, 2, 3, 4}.