RE: Variables and HTML

Thanx Michael, that makes a lot of sense to me. 

Cheers,
<prs/>

-----Original Message-----
From: Michael Kay [mailto:mike@xxxxxxxxxxxx] 
Sent: Viernes, 11 de Marzo de 2005 12:47 p.m.
To: xsl-list@xxxxxxxxxxxxxxxxxxxxxx
Subject: RE:  Variables and HTML

> That leaves me with the question why it is not encouragable to couple 
> transformers and serializers... may we assume that serializers are 
> kept out of the spec's domain is because serializers are too system 
> specific?

It comes down to pipelining, or closure.

A property of XSLT is that the input data model is the same as the output
data model. The operations in XSLT take trees as input and produce trees as
output. The language is "closed" over the data model. The benefit of this is
composability: any two transformations can be combined to produce a larger
transformation. Hence pipelines.

Serialization should be separate because it breaks away from the data model
and produces something different: its output is a different kind of thing
from its input. Only by keeping serialization separate from transformation
do you preserve the closure property of the transformation language, and
hence its composability.

Michael Kay
http://www.saxonica.com/