Re: streamable xpath subsets

Pratik's document discussing limiting XPath productions allows only
element nodes to be in the output node set. As I mentioned in this
week's meeting: "It seems to me that a transform that results in a
single text node should be supported. For example, an app stores binary
data as base64 in an XML element and wants to hash (on signing and
validation) the original raw binary. On validation, use XPath to select
the text, then feed that to the base64-decoding before hashing.". I
would suggest adding the ability to return a text node.

More comments...

Why isn't the id() function allowed? Could you please explain why that
is not streamable?

I've also been looking on the Web for research about efficient XPath
streaming and have found a number of works including some describing
very efficient streaming XPath algorithms that do not require many of
the limitations proposed in Pratik's document. Here are some links:

http://www.pms.informatik.uni-muenchen.de/publikationen/PMS-FB/PMS-FB-2001-16/slides_dagstuhl_2002.pdf

http://cs.nyu.edu/~deepak/publications/icde.pdf

http://xml.coverpages.org/BartonPLANX2002.pdf

To me, the above articles suggest that many of the XPath limitations
proposed in Pratik's document are unnecessary wrt streaming and
efficiency. Comments?

Ed


On Tue, 2009-09-08 at 12:07 -0400, Frederick Hirsch wrote:
> I uploaded  a version of the Streamable XPath Subset document Pratik  
> sent, so that colors are preserved on the web
> 
> It is available at
> http://www.w3.org/2008/xmlsec/Drafts/proposals/Streamable-XPath-subset.html
> 
> I created a "proposals" directory in CVS for this sort of thing.
> 
> Scott, perhaps you can include this link in the minutes so readers  
> know what we were discussing.
> 
> regards, Frederick
> 
> Frederick Hirsch
> Nokia
> 
> 
> 
> 
>

Received on Wednesday, 9 September 2009 23:37:35 UTC