It really doesn't matter whether Altova's XSLT processor is fast or slow. Until they fix their parser so that whitespace-only nodes in mixed content are not thrown away, Altova's product is useless for documentation work. As it is, if you have two adjacent words in a paragraph that are each individually tagged, Altova throws away the space between them. I sent them a message about this over a year ago. They incorporated the example I gave them into their docs to explain the problem -- but they have not fixed it in two major releases.