Originally posted on the main site.

This document describes how schema extensibility is typically achieved in commercial enterprise software applications; in particular, in customer relationship management (CRM) applications. A future update will discuss newer extensibility mechanisms such as native XML column types in databases.

(more…)

So here is a typical data model pattern:

create table master ( mid int not null auto_increment, … , primary key (mid) );

create table detail ( mid int not null, did int not null auto_increment, … , primary key (mid, did) );

( detail.mid is a foreign key to master.mid but no one uses these in MySQL… )

So this works great if you’re using the default MyISAM storage manager but it does not work using the InnoDB storage manager (RTFM!). Unfortunately, MyISAM does not provide ACID properties which means you should not be using MyISAM unless you have another way of recovering from data loss. My data is valuable so I have to use InnoDB, which means I have to restructure the detail table so that detail.did is the primary key and detail.mid is just a field with an index.

I really wish MySQL had one storage manager that actually worked like a relational database is supposed to, instead of several half-working storage managers.

Links to SQL Server Tuning Info:

Index Tuning Wizard SQL Server 2000 (MS Technet)

Index Covering Boosts SQL Server Query Performance (devx, basic article)

Optimizing Database Performance Overview (MSDN/SQL2K online manual)

Red Gate Software SQL Server Performance

Which is better? Jetty or Tomcat? Both are open-source servlet / jsp containers availabler under the apache license. Both implement the latest servlet and jsp standards. Technically they are neck and neck.

In the end I decided to use Tomcat because it has better documentation. Jetty’s manual sucks.

The Economist ran an article on the Cell chip so I figured I better check it out. Nicholas Blachford has an excellent review of the technology here. High level summary: a PowerPC with 8 vector processor units, no virtual memory, and no caches. It was designed for media-intensive consumer-electronics applications like gaming consoles and HDTV. According to the Blachford article, the Sony PS3 will have 4 Cell cores and the graphics will probably scream.

Blachford also speculates that cheap desktop and server computers based on the Cell chip may finally beat out the x86 architecture for general purpose computing. His argument is that Cell will deliver more processing power at lower prices (and at a lower temperature), so industry will naturally start to use Cell parts for general purpose computers. He also speculates that Cell will become the de factor standard architecture for the gaming software industry: Cells are so cheap that you could ship a PS3 as a PC daughtercard; the availability of a $100 PS3 that one could embed in a PC would increase the attractiveness of the PS3 software platform and hurt the PC game platform.

Which brings us to the key point about Cell: it’s going to be a bitch to write software for this thing. Vector processing has been around forever. It’s great for applications that are, well, vectorizable and it blows for everything else. The ability of Cell to win the GP computing war against x86 will depend on how well it can run modern software, with object-oriented programming, virtual machines, and 14 zillion levels of abstraction to help you forget that the software actually runs on grody stuff like hardware. It will be interesting to see if IBM has a research project to get a fast JVM or CLR on Cell. You can think of some obvious ways to partition the problem, like a separate JVM per cell or threads from a single JVM on a set of cells, but it’s not clear whether the memory architecture will make this worthwhile.

Pointer to a good review article on business rule engines; appeared in Infoworld in June 2004.

Infoworld Business Rules Engine Review

The article forgot the unfortunately named but open source DROOLS engine.

More to come…

Currently spending time investigating XML databases. Most XML Databases use the XML:DB api which is defined by the XML:DB Consortium. Their site (http://xmldb.org) is down. Anyone know what happened?

In one sense, this might be a good thing. There are too many open source XML database implementations so some consolidation is in order. I’d like to see the Jakarta Xindice guys give up and join the Exist project. I’m migrating one of my apps from Xindice to Exist, mostly because Xindice seems to have joined the ranks of the living dead; but also because I’m embarassed to say “Xindice” in public. I can’t say it without an Italian accent and it makes me feel like Kevin Kline in “A Fish Called Wanda.”

So I decide to spend the day figuring out if MDA (Model-Driven Architecture) tools, in particular AndroMDA, are ready for prime time. It was about as much fun as a stick in the eye, and took a lot more time.
(more…)

Current instructions for configuring Tomcat for JAAS-based Windows authentication.
This is the solution for my previous post.
(more…)

It started out as a simple challenge: figure out how to configure a Java Servlet app so that you can authenticate using your Windows password. My app uses container-managed security in Tomcat so I figured it would be trivial. Ha ha.
(more…)

I’m using docbook to write manuals for some software that I wrote. For old timers, think of docbook as an XML version of LaTeX. Instead of processing the files with latex and dvi2ps, you use XSL programs to transform the XML into HTML, PDF, etc. I also wanted text formats so I could store the docs in source control and process them using ANT.
(more…)

The (US) Federal Trade Commission issued recommendations to Congress on ways to reform patent law. The whole affair was rather disappointing. Most of the testimony on software was from people with a vested interest in the status quo: intellectual property lawyers and people from big software companies with large patent portfolios.
(more…)

Havoc Pennington has apparently shaken up the GNOME community with his essay on why GNOME should move to Java. I think the essay is great but it ignores the most important reason for switching to Java: winning.

(more…)

No suspense today. Hotspot’s dirty little secret is that it doesn’t work. Hotspot has had a bug for years somewhere down in locore that causes it to crash hard under load. The only dependable workaround is to disable Hotspot (run the JVM with Hotspot disabled using “java -Xint”). This impacts most versions of Weblogic, Websphere, JBoss, Tomcat, etc. I know of at least one commercial business application that had to ship with Hotspot disabled because of this bug.

http://developer.java.sun.com/developer/bugParade/bugs/4724356.html

Java is going to fail because it ignores the most important segment of the market. We agonize over J2EE vs .NET without realizing that, from a business perspective, it doesn’t matter. Both platforms have the same technical characteristics: compiled languages running on a virtual machine, heavy and expensive deployment requirements. Where it really counts — cost to deploy and programmer productivity — the two platforms are nearly identical.

(more…)

I read in the FT today that the Senate passed a bill to restrict offshoring. I think the offshoring thing may blow over in the next couple of months since:

• Offshoring is essentially a long-term bet on a strong dollar. If the dollar continues to weaken (and if it floats against pegged currencies like the yuan) then offshoring becomes much less attractive and much more risky financially. Imagine if all your savings due to offshoring suddenly turn into an expense-line risk due to currency fluctuations.
• We will eventually adapt to a post-offshoring economy. The high value-add sector of our economy is going to be in biotech and nanotech, not software or semiconductors. People will eventually get this and start retooling themselves for this new economy
• Offshoring will in fact make the US economy more productive. This study argues that many sectors of the economy (retail, healthcare, etc) can now be automated because IT will become much cheaper for them due to offshoring.

Had a lot of “fun” today figuring out why my web app would not startup when my laptop was not connected to the network. I’ve known about this problem for awhile but I ignored it. Tomorrow I’m planning to demo said web app to some friends over lunch so I finally needed to fix it.
(more…)

Business Week Online reports a Boston-area company that was looking at sending work to India. They were quoted $40,000 per year per programmer. They decided to try to hire U.S. programmers for $45,000 per year and the labor market is so bad that they found takers. This is the inevitable consequence of offshore competition for programmers. Gone are the days when a newly-minted CS major with 2 years experience could expect to make $80,000 per year and get a signing bonus and a Senior Developer title. We should eventually expect the price of real estate and cars to drop as well as standards of living take a hit.
(more…)

I was looking for a UUID implementation in Java and found this: JUG — Java UUID Generator

It’s true to the algorithms in the Internet Draft covering UUID generation. One surprise: name-based UUID’s don’t work the way I expected them to; be careful out there. Unfortunately, reading the internet drafts themselves are as exciting as watching paint dry. Luckily, you can just grab the code and forget about how it works.

I’ve been working with Java / XML binding tools for the past couple of years and I finally found one I like. In general, these tools have a compiler that takes an XML document definition (XML Schema or DTD) and generates classes (Data Transfer Objects for you Pattern geeks) corresponding to the definition. They also include a runtime marshalling framework for converting between XML documents and objects.
(more…)