View Single Post
  #2 (permalink)  
Old December 2nd, 2006, 11:46 AM
mhkay's Avatar
mhkay mhkay is offline
Wrox Author
Points: 18,438, Level: 59
Points: 18,438, Level: 59 Points: 18,438, Level: 59 Points: 18,438, Level: 59
Activity: 0%
Activity: 0% Activity: 0% Activity: 0%
 
Join Date: Apr 2004
Location: Reading, Berks, United Kingdom.
Posts: 4,954
Thanks: 0
Thanked 290 Times in 285 Posts
Default

>Why was these design chosen?

I'm always a bit cautious about answering questions of historical fact, I'm not a historian and it's very easy for conjectures to turn into urban myths.

There have been attempts to design transformation languages that process a document sequentially, for example STX and XStream. Generally they are harder to use or have more limited functionality. Doing something like building an index or a table of contents when you haven't got the ability to navigate anywhere you like in the source tree isn't easy.

Some people also argue that a good optimizer should be able to detect those XSLT stylesheets where serial processing is possible and avoid building the tree. Personally, I'm sceptical.

Michael Kay
http://www.saxonica.com/
Author, XSLT Programmer's Reference and XPath 2.0 Programmer's Reference
Reply With Quote