Using SQL Server 2005, there's an interesting and really fast way to do it. You should read all the explanation and be sure to understand it before using it. You should test that it works correctly before full implementation.

The explanation is here:

http://www.sqlservercentral.com/articles/T-SQL/68467/

And the example is here:

WITH SampleData(Pos, item, SID) AS(

SELECT 1, 'tire', 1 UNION ALL SELECT

NULL, 'black', 2 UNION ALL SELECT

NULL, '50 $', 3 UNION ALL SELECT

2, 'car jack', 4 UNION ALL SELECT

NULL, 'blue', 5 UNION ALL SELECT

NULL, '35 $', 6 UNION ALL SELECT

3, 'screwdriver', 7 UNION ALL SELECT

NULL, 'red', 8 UNION ALL SELECT

NULL, '3 $', 9)

SELECT *

INTO #Samp

FROM SampleData

CREATE CLUSTERED INDEX SampID ON #Samp(SID)

SELECT *

FROM #Samp

DECLARE @Pos int, @anchor varchar(100)

SELECT TOP 1 @Pos = Pos

FROM #Samp

ORDER BY SID

UPDATE a SET

@Pos = Pos = ISNULL( Pos, @Pos),

@anchor = item

FROM #Samp a WITH (TABLOCKX)

OPTION (MAXDOP 1)

SELECT *

FROM #Samp

DROP TABLE #Samp