As Paul Flessner publicly announced at his TechEd 2005 keynote, the world-wide RTM launch date has been set to the week of Nov 7th 2005 and the June community tech preview (CTP) of SQL Server 2005 has been released. Again, we have added improvements in stability, performance and scalability. And like for the previous CTPs, I have a list of some important changes in the XML area. Note that the release comes with a more general breaking changes list that you should take a look at as well.
Here is the list of changes that may affect existing usages of XML:
- As previously discussed, the XML data type methods cannot be used directly in CHECK constraints anymore but need to be wrapped in a UDF
- Whitespace parsing rules for attribute values have been fixed
- Whitespace preservation measures in serialization: The last whitespace character in a preserved whitespace only text node will now always be entitized. This guarantees that replication and BCP out/in preserves the whitespace information. An option to not entitize it will be added in a future CTP.
- Some more XQuery alignments were done:
Several string functions are now aligned to the July 2004 and later XQuery function and operator working drafts regarding their behaviour when the empty sequence () is passed:
- fn:substring() returns the zero-length string ('') instead of () when any of the parameters is empty sequence.
- fn:string-length() returns 0 instead of () when empty sequence is being passed.
- fn:contains() treats () input as '' (the zero-length string).
- replace value of with an empty sequence as the new value now raises a runtime error unless the empty value is given explicitly as (). This guarantees that empty sequences that are generated due to our mapping of XQuery runtime errors to the empty sequence will not accidentally overwrite values.
- Additional XSD restrictions:
We disallow xs:ID/xs:IDREF/xs:IDREFS types on elements. The semantics was not completely implemented in earlier CTP and Beta versions, so we decided to disable them in this release instead of giving incorrect behaviour.
In addition, we still have some additional work to get done regarding the above mentioned whitespace preservation options and some encoding issues in the upcoming CTPs. As always, please use the new CTP, and provide us feedback. We have done lots to improve performance and appreciate feedback on where you want us to continue to improve (even for after RTM :-)).
Here are my quick notes from Paul Flessner's keynote:
- The Daily Show's Sammantha Bee's technical benefits translator translates the three key benefits of SQL Server 2005 as follows:
Availability and reliability -> Downtime is for suckers
Security -> Hi Hackers … bite me!
Scalability and Performance -> SQL Server 2005 is like spandex pants (it fits any size)
- Paul announced performance figures:
TPC-C: 37% better perf at 17% lower cost than SQL Server 2000 numbers on HP 64-P Itanium hardware using scale up (hitting 1082203 T/S; 7% better than Oracle at 37% less cost)
TPC-H 1TB: 162% better at 54% less cost (on same platform again faster and cheaper than Oracle numbers)
He also noted that these numbers are somewhat beyond most real-world requirements.
- Failover demo'ed with battle robot chopping the network switch of the 32-bit SQL Server 2005 installation and showing how the 64-bit installation absorbed the additional, failed-over load.
- Reporting Services will be available in all editions (even Express).
- The world-wide launch date is now officially set to the week of Nov 7th! The June CTP is out now!
- While SQL Server is 3rd in revenue share (Gartner: 20% 2004), but has the highest unit share numbers (IDC: 41% in 2003, and 34% in enterprise market).
- Cost Chopper Contest: www.sqlserverchopper.com Migrate from Oracle and win a custom chopper.
- Paul announced a free downloadable migration assistant tool to migrate from Oracle to SQL Server…
I am happy that we now have the launch date fixed. I agree with Paul that especially the TPC-C numbers are getting ridiculously outside any real-world use case. The industry should better focus on solving real performance issues.