Enjoy Every Sandwich

Thoughts on SQL, XML, .NET and sometimes beer.

<October 2008>
SuMoTuWeThFrSa
2829301234
567891011
12131415161718
19202122232425
2627282930311
2345678


Navigation

Tools

List O'Links

Kent's Other Stuff

Subscriptions

News

Please read these
Notices and Disclamiers

Post Categories

Article Categories



Monday, August 30, 2004 - Posts

Aakash Kambuj has started a blog

Aakash Kambuj, a developer on the SQL Server 2005 team, has opened a blog on MSDN. He's promising us content on “the more esoteric features” of SQL Server 2005. He's got some good jokes here too.

SUBSCRIBED.

posted Monday, August 30, 2004 11:24 AM by ktegels

Depreciating the SOAP formatter

Well, I hate to say they “they told us so.” But they did. A shot has been taken across the bow of Remoting with SOAP. It's interesting to see how the camps are dividing up. There's a lot folks (notably MS employees and the influentials) that are cheering this. Then there others who've got a lot to lose with it railing against it, like Dale Brenner:

Such a move essentially destroys my investment, and give great creedance to the open source argument.Every day I battle those folks who advocate open source. How do I maintain any credibility with my customers, if it appears that you willing to arbitrarily deimplement critical features... To be clear, the move you are contemplating would go a long way toward driving me out of business.

While I couldn't really give a bee's knee about this issue personally (I abandonned Remoting altogether after PDC), I think it says something that Microsoft is willing to put a stake in the group in such a public forum and see what the feedback is. Now seems like a good time to say “2.0 is last iteration of it“ and then be done with it, though.

Let's be fair to MS here too: What prevents an OSS project team from depreciating feature? Sure, with OSS you can still with an old version all you like and try to work around its issues and spend your own effort trying to keep up with the base product if you like. Do mod_log_referer and mod_log_agent ring any bells? The situation with .NET isn't that different provided the OS allows you run different runtimes side-by-side. Of course with .NET, your code should just keep working.

I think more importantly though, such a move gives Indigo more urgency. It seems like something better than remoting and ASMX and we need it before Longhorn ships.

 

posted Monday, August 30, 2004 10:18 AM by ktegels

Don't build-up XML instance types serially!

I'm half temped to start a "museum of dumb ideas I've had". Here's an example using YukonSQL Server 2005. This is why do not want to construct large XML documents in a serial fashion using the XML type...

declare @doc xml
declare @index int
declare @started datetime
declare @ended datetime
declare @order tinyint
declare @interations int
declare @dur float
set @order = 1
while @order < 33
begin
 set @doc = '<root />'
 set @doc.modify('insert <node id="{sql:variable("@index")}" /> as last into (/root)[1]')
 set @started = GetDate()
 set @interations = power(2.0,@order)
 set @index = 0
 while @index < @interations
 begin
  set @doc.modify('insert <node id="{sql:variable("@index")}" /> as last into (/root)[1]')
  set @index = @index + 1
 end
 set @dur = datediff(ms,@started,GetDate())
 select @order,@dur 'duration',@interations 'interations'
 set @order = @order +1
end

Run it. First three people that report their execution times for any iterations above 24 get a free gmail invite.

posted Monday, August 30, 2004 7:55 AM by ktegels

Stamping Out Outer-Joins (sort of)

Christian Kleinerman, one of the SQL Engine PMs, posted this to various Yukon newsgroups this weekend.

For years we’ve provided non-ANSI syntax to perform outer joins. (aka " *= outer joins ", "old style outer joins"). This syntax is ambiguous under some circumstances, which makes it hard to predict what the intent of the query is, let alone guarantee the desired results L

The ANSI standard provides non-ambiguous syntax to perform outer joins ( { LEFT | RIGHT | FULL } [ OUTER ] ) which has been available in the last few releases of SQL Server.

Given the above, we’ve often communicated our intention to deprecate this syntax in a future release of the product.

We’re considering disallowing old style (*=, =*) syntax in SQL Server 2005 under database compatibility 90. It would still be available for backwards compatibility under 80 and earlier compatibility levels (sp_dbcmptlevel). Applications would need to be changed to ANSI compliant syntax before moving to 90 compat.

Please let us know if you see any issues or concerns with this plan. All feedback welcome.

I for one have no problems with this and think its a step in the right direction. If you disagree, feel free to leave a comment here or send me a message -- I'll fowarded them up to Christian.

posted Monday, August 30, 2004 6:24 AM by ktegels




Powered by Dot Net Junkies, by Telligent Systems