On Fri, 28 Jun 2002, Tom Bradford wrote:

> On Friday, June 28, 2002, at 10:06  AM, Fernando Padilla wrote:

> The client and the backend can assume this based on the API calls that 
> are made.  getDocument implies a document resource, while (in the 
> future) getBinary implies a binary resource.  The server will kindly 
> throw exceptions if you make a call that the collection won't be tasked 
> for.

-  wait, so the client code will return an XindiceResource which will 
implement both BinaryResource and XMLResource???????  I wouldn't be too 
happy about that.

-  as an aside, I could see an XMLResource implementing the get/setContent 
methods to serialize/parse the binary content into SAX/DOM then call the 
get/setContentAsSAX ( actually this should be an easy patch )



Essentially though an interim solution where the Resource Type is stored
at the Collection level is great!, I still think we should start moving
things over so that this information is maintained at the Resource
level... This might break things ( like you said, it might affect the
token stream or something ), but it will make things better in the long
run.  The only thing I see is that it will probably break more than we
expect because most of the code probably has an assumption of only
XMLResources ( like possibly the XPath/Indexer )....


What do you think?


fernando


Reply via email to