XML 2004: Late short trip report and links to my presentation
It has been more than a month already, since I attended the XML 2004 Conference in Washington DC. My presentation on XQuery in the context of relational databases was well-attended. Attendees had the unique opportunity to compare the (not too diverging) philosophies of Oracle and Microsoft on this topic since my talk was in the same session as Stephen Buxton's talk on SQL, SQL/XML and XQuery. I was just a pity that a scheduled IBM talk on the same topic got cancelled (we could have found out first hand what their plans are).
You can find my submitted paper at http://www.idealliance.org/proceedings/xml04/abstracts/paper254.html (follow the links for the HTML or PDF version). Note that at the time of this writing, it is not the latest version that contains some minor corrections and reformulations. The presentation can be found at
Here are my short impressions:
- The hot topics in the areas where I was hanging around were (not surprisingly):
- Binary XML: a good or bad idea?
- XQuery: more than a fata morgana?
- XML and databases
- As Len pointed out: Many good people attended. However I think the conference will loose some of its value and raison d'etre over the next two years. As XML becomes more entrenched, the technical challenges are being solved, and no new XML core technologies are being designed and marketed anymore, the organizers will have to find a new focus for the conference (besides being a cool place to meet interesting people :-)).
- Congrats to Jean Paoli and Robin Cover for their deserving win of the XML Cup.
Here are some more details on the three topics above.
Binary XML: a good or bad idea?
My former boss and friend Rich Rollman (now with AgileDelta) organized a townhall meeting on this topic. Since I was busy on the XQuery townhall meeting at the same time, Soumitra “represented“ our position. I've heart it was some times quite hot in that room :-).
As I said in an earlier post: I am not against using binary encodings for XML in closely-coupled environments. But having a single standard for a binary XML encoding is just a bad idea. It is similar to making Shift-JIS5 or Windows-1252 a mandated encoding for XML without concern for other requirements that are not satisfied by these encodings. If a certain user community needs a more efficient encoding, it should define its own encoding that solves their performance need (and please just put it into the XML declaration encoding property and not the MIME type!). However, that encoding should not be used in loosely-coupled environments unless it really gives 2 to 3 order of magnitude improvements. I would expect that there would be at least about 3 to 5 such binary encodings that will find traction and that cater for different optimizations (I just hope they will stay in closely-coupled environments, but given the push from some of the mobile phone community, I fear that they will not)...
XQuery: more than a fata morgana?
I already mentioned that I (together with Dana Florescu, Jonathan Robie, Ron Bourret and Michael Kay) sat in an XQuery townhall meetings. We had some interesting discussions about why XQuery is taking that long and what an XML database really is (see below for a one-liner) and why we need this over SQL and relational databases. The audience was an interesting mix of standard people, tool vendors and “real” users.
On the presentation side: Besides Stephen's and my talk, we also had DataDirect “spamming“ (or was that “egging“? :-)) the conference with several product presentations by Jonathan Robie on their upcoming mid-tier XQuery and SQL/XML engine support. All these presentations bascially communicated the message:
- Natively storing XML documents is an important, scenario-enabling technology for general database systems,
- SQL/XML and in general, relational XML functionality is important, but is not in competition with XQuery but provides synergies with XQuery,
- XQuery is the language to query into and across stored XML documents.
There was also an interesting presentation by Borkar (based on work done with Mike Carey and Dana Florescu at BEA) on how to extend XQuery with grouping (and more direct outer join operators). I am sure that the XQuery WG will take a look at this functionality for the next version of XQuery.
The number of XQuery related presentations and the attendance of all the XQuery events (including a Tutorial by Priscilla Wamsley) made it clear that there is interest in the technology and products. Time will tell if they will deliver what the users need.
XML and databases
In general, there were less so-called native XML DB vendors present than during earlier conferences. Georg Feinberg (Sleepycat) gave a general presentation on "native" XML databases: What are the issues, how to store it etc. Not very deep for actual implementers but interesting for people that are new to the area. MarkLogic showed an interesting demo build with their XML database as back end and XQuery during one of the product talks. Very slick.
In my opinion, "native XML DBs" like "native OODBMS" will be relegated to a specific niche as the big relational database vendors add native XML capabilities to their databases. A native XML database in my definition is a database, that can store XML in a way that preserves the infoset-fidelity while providing a declarative query language and update language over the stored XML while providing the general database management functionalities such as backup/restore, durability, correctness, concurrent access etc.. Based on this definition, obviously Oracle 10g and SQL Server 2005 provide native XML database capabilities (and more beyond that).