RE: Xpath across a network

[Christopher Raber]

Title: RE: Xpath across a network

Brad,   In the past I worked for an object database company (GemStone). We had both Smalltalk and Java versions of our products. Many customers would start out trying to access their object model over the wire in its “pure” form (i.e. as if the remote objects were in local memory). What ended up happening 9 times out of 10 is that the access to said objects was not very efficient. Typically way more data than needed was transmitted to the client.   What was typically done to rectify this is moving to a service based interface where the server would publish a “coarse grained” interface to the client, and return a small network of objects to the client by value. This requires more design work to carve up the client server interface, but almost always is way more performant.   XML/RPC seems to be a good way to model said coarse-grained interfaces. Yes, this is more work than just going at your XML docs using XPATH! But I don’t think it is reasonable to think that the latter is going to work well over the wire…   Regards,   -Chris.   Chris Raber, Systems Engineer, AvantGo Inc. http://www.avantgo.com/   -----Original Message----- From: Brad O'Hearne [mailto:[EMAIL PROTECTED] Sent: Monday, September 24, 2001 4:32 PM To: Xalan (E-mail) Subject: RE: Xpath across a network   Chris,   Thanks for the suggestion.  I have been pondering something similar, but I have a couple of issues:   1) I don't want to recreate a query language if I don't have to.  Xpath meets all requirements -- I just need to de-couple the results from the "mother ship" as mentioned prior.  Now changing the result of the xpath query to XML before returning it across the wire would be possible.  I am looking for a method on the Node, NodeList, or NodeIterator interfaces that will create xml out of the contents...but am not finding any.  I suppose this will all have to be done manually?   2) I would prefer not to propogate a parsing requirement to my client, rather move an object structure (great thing about RMI).  The NodeList or NodeIterator would be fine, just as long as they were detached from the source -- I do not want to push more data across the wire than need be.  Is there a way to detach these?  Surely others out there have a similar requirement...   BradO -----Original Message----- From: Christopher Raber [mailto:[EMAIL PROTECTED] Sent: Monday, September 24, 2001 11:36 AM To: '[EMAIL PROTECTED]'; 'Xalan (E-mail)' Subject: RE: Xpath across a network Instead of an XPATH interface, how about an XML/RPC based service? The incoming request could be a XML structure describing the query, and the result could be the XML tree that matches and nothing more. -----Original Message----- From: Brad O'Hearne [mailto:[EMAIL PROTECTED]] Sent: Monday, September 24, 2001 2:24 PM To: Xalan (E-mail) Subject: Xpath across a network Hello all, I asked a question not too long ago that was up this same alley, but I have a more direct question that perhaps you can help with. I have xml data on a server, stored as a DOM, and I need to expose to a client the ability to query the document.  XPath seems the perfect interface.  However, I am concerned that if a client performs an XPath query, the resulting data returned won't be only the nodelist, but a whole bunch of other unnecessary info and data structures that the nodes reference (like potentially the entire DOM).  I am not totally sure about that, but my last post about this suggested strongly that resulting nodes from an XPath queries had underpinnings to the "mother ship", if you will. So...any advice on how to allow clients to query a remote DOM, and receive only the actual results across the wire? Thanks in advance! BradO ----------------------- "It's such a fine line between clever and stupid."  -- David St. Hubbins, Spinal Tap ----------------------- Brad O'Hearne Software Scientist Irvine, CA  92618 949.743.5121 Tel 949.753.7434 Fax [EMAIL PROTECTED]