xquery & web server in browser

[Robert Koberg]

On Jun 18, 2009, at 4:24 PM, John Snelson wrote:

> Florent Georges wrote:
>> Robert Koberg wrote:
>>> I was talking about the ability for the XML DB vendors finding it   
>>> easier to have XQuery run natively in their DBs as opposed to XSL.
>>  Yes, but I wonder why XQuery would be more efficient than XSLT  
>> "natively on the server"?
>>  I would expect them to be as efficient as the other, excepted that  
>> XQuery module system could ease compilation, and on the other hand  
>> XSLT facility for grouping and pattern matching that should  
>> benefits runtime for those features (opposed to writing them by  
>> hand in XQuery.)
>
> The problem is the dynamic dispatch in XSLT - you can't know at  
> compile time which template will be invoked where. This is a problem  
> because:
>
> 1) You generally can't get accurate static type information across  
> template application.
> 2) You generally can't optimize across template application  
> boundaries.
>
> I think these aren't necessarily as much of an issue has they have  
> traditionally been thought as. On the one hand we are seeing a  
> number of schema-less XML databases, which rely less on static  
> typing. On the other, I think that most XSLT stylesheets wouldn't  
> benefit all that much from optimization across template application  
> boundaries anyhow.

What XSL would benefit from is not having to load each document into  
memory. Ideally I would use XQuery to gather the requested parts of  
various documents in the existing structure/schema. Then, allow XSL to  
transform the XQuery builtup nodeset without having to create some  
DOMish (even if it is the XSL processor's optimized DOM) object.

best,
-Rob