[xquery-talk] Comparing XML Databases to Object Databases

[John Snelson]

Michael Kay wrote:
> The database market is incredibly conservative and risk-averse, both in
> relation to new technologies and new suppliers. MySQL and PostgreSQL have
> shown that you can break the mould in terms of suppliers (but only by
> cutting the price to zero), but no one has yet proved it in terms of
> technologies. 

But interestingly, object-SQL mapping technologies like Hibernate mean 
that there are huge numbers of developers who never have to touch SQL or 
any of the advanced features of the database. The less you are invested 
in the back-end, the less you are likely to care if it changes.

> The advantage of going with a hybrid database is that you are effectively
> deferring a decision on your technical direction, which makes it an easier
> decision. The disadvantage is that you end up leaving junior programmers to
> make the real technical decisions on a case-by-case basis, and they will (a)
> get it wrong half the time, and (b) fight with each other over which way is
> best [yes, I've watched this happening, and it's painful]. But that argument
> won't sway decision-makers who think they can have their cake and eat it,
> and who often have a dismissive attitude to arguments based on technology
> anyway.

I think open source is starting to disrupt this process, because 
decisions can increasingly be made on technical merits before the people 
in charge of the money need to get involved [1].

> The biggest weakness in the object database market was standards. The
> companies collaborated on ODMG, but they implemented different parts of it
> and the products were never interoperable. Also, the technical architecture
> was a bit of a mess - because of the focus on persistent C++, you ended up
> with a navigational DML and a query language that bore very little
> relationship to each other, and with applications that were very closely
> tied to the physical data structure. So unless you needed the raw speed, for
> example in CAD applications, there were few benefits. I think that the XML
> database scene is very much better served in this respect.

I agree.

> But however good the technical merits, there's no doubt that XML databases
> have an uphill struggle convincing corporate decision-makers, especially
> when the Oracle/DB2/Microsoft sales people are saying "we can do it too, you
> don't need to take any risks, you don't even have to decide yet what your
> technical direction is going to be." Many people prefer to stick with the
> devil they know. Some don't even recognize that they are dealing with the
> devil...

I suppose I can't really comment on incumbent RDBMSs and "the devil"... :-)

John

[1] http://blogs.sun.com/webmink/entry/the_adoption_led_market

-- 
John Snelson, Oracle Corporation            http://snelson.org.uk/john
Berkeley DB XML:        http://www.oracle.com/database/berkeley-db/xml
XQilla:                                  http://xqilla.sourceforge.net