At nerd camp

My wife got an email from Nate‘s wife with a picture of me at the Flex Bugquash.   Thankfully, without an identifying caption.

That’s my rickety old Gateway laptop  sporting Windows Vista.  I think one other person there had an HP with Windows, but otherwise it was wall to wall Macbooks.  Unfortunately, that meant I spent a lot of time troubleshooting problems without tech support.  Wifi didn’t work for me either at the Adobe campus, once all the Macs showed up.  Apparently there’s a prioritization.

Anyway, there were lots of snacks (I had 3 times my share of M&Ms), lots of hacking, and while I didn’t get a patch submitted to fix a bug, I learned how to hack the Flex SDK, and got some unit tests around DateFormatter running.

I’m now a lot more comfortable in Flex and am incorporating what I learned with Sprouts and Fluint.  I finally feel comfortable enough to work on automating a Flex build & test suite.

Using Flash 10 debug projector in Flexbuilder 3.3

Here’s how I set up Flexbuilder 3.3 to use the flashplayer_10_sa_debug.exe.

(I still haven’t figured out how to get it to launch full screen, and you may need a commercial app that wraps projector to do that — though it’s probably possible with the open sourcing.)

Download the debug standalone flashplayer 10 here:

or whatever the current version is from:

Locate a swf and lauch it directly, choosing flashplayer_10_sa_debug.exe when prompted by windows for the application to use.  This will set up the Windows registry to associate the player with .swf files.

Set up your run configuration in Flexbuilder to open the project .swf file at c:\path\to\YourProject-debug\Whatever.swf (instead of http:\\localhost\YourProject-debug\Whatever.html)


This is a bit of a hack, since it needs done for every project, but oh well.

Now it will launch, but you’ll get a pesky alert like:

Flex Builder cannot locate the required version of Flash Player.  You might need to install Flash Player 9 or reinstall Flex Builder.

Do you want to try to run your application with the current version?


To get rid up this and use Flash 10 (and the new Flex SDK), first download the SDK:

or get whatever the latest version is from here:

To have Flexbuilder use the latest SDK, open Windows > Preferences > Flex > Installed SDKS


and add the downloaded SDK, setting it as default.

Open the Flex build path, click on the Library Path tab, and delete playerglobal.swc from Flex 3.3.  Then click “Add SWC” and locate the playerglobal.swc in  path/to/new/sdk/frameworks/libs/player/10


Then specify the flashplayer included in the SDK download for the project (again, a hack that needs repeated for each project).  Right click on the project, select “Properties”, and select “Flex Compiler”


Then enter 10.0.0 in “Require Flash Version”.  This should work.

See also the following links, which helped me out:

Also see:

to use the HFCD which pre-loads the compiler in a daemon to speed up compile times.

And for the Zinc Flashplayer laucher which can resize the projector on launch (among other things.)  The price is a bit steep though ($300), if that’s all you want to do.

And here’s a blog post about setting up a free Flex development environment:

Flexbuilder and SVN

I almost thought I had something there.  Here’s what I did:

Attempted to install subversive from:

Got the following error message:

Subversive SVN JDT Ignore Extensions (Optional) (Incubation) (0.7.7.I20090224-1900) requires plug-in “org.eclipse.jdt.core”.

Installed Java Development Tools from Eclipse software updates:

Europa Discovery Site
Java Development
Eclipse Java Development Tools 3.3.2r33x_r20080129-..

This hung while using IBiblio.   I restarted FlexBuilder and tried again.  This time using xmission.  Much better, and successfull.  FlexBuilder rebooted and then I went to the polarion update site above for Subversive and successfully installed.  Rebooted again, and right click on my project:

Team > Share Project

Entered my credentials, but then, bam:

Share project was failed.
Share project was failed.
Selected SVN connector library is not available or cannot be loaded.
If you selected native JavaHL connector, please check if binaries are available or install and select pure Java Subversion connector from the plug-in connectors update site.
If connectors already installed then you can change the selected one at: Window->Preferences->Team->SVN->SVN Client.

Selected SVN connector library is not available or cannot be loaded.
If you selected native JavaHL connector, please check if binaries are available or install and select pure Java Subversion connector from the plug-in connectors update site.
If connectors already installed then you can change the selected one at: Window->Preferences->Team->SVN->SVN Client.

The duplication is odd, but I don’t know if it’s symptomatic of anything.

U can’t believe FlexBuilder doesn’t have support for subversion, but I can believe Eclipse is borked enough to fail on something trivial like this without any help from Adobe.

My first Flex bug

I just logged my first bug against Flex:

I’m not sure if it’s actually a problem with the compiler or with FlexBuilder.  My guess is that a compiler bug is not properly handled by the UI — which makes it hard for me to track down, since the UI problem may be masking the underlying issue.

No doubt, the root cause is some bad coding on my part.  But bad code shouldn’t be able to break the compiler, and a compiler failure shouldn’t be able to break the UI.  To be more specific, it doesn’t break the UI, but it seems that there’s no way to correct the error, recompile and continue.  Once it breaks, I need to restart FlexBuilder to be able to recompile.  But if the code that triggers the compiler bug isn’t fixed, it’ll come back up on restart.  Not very useful for debugging. I suppose I should know what I did wrong and just fix it, but if developers knew everything you wouldn’t need compilers.

Anyway, I’d probably scrap Flex right about now, and mutter something about half-baked betas released to production, but because Adobe is having a Bug Quash next weekend, and I’m going, I’ll try to figure out more information about it and take it with me.

I’ll be doubly impressed if I get more than a “can’t repro” or “as designed”, though I’d be impressed enough if someone actually look at it at all.

I won’t bother posting my project that triggered it, though I hope to be able to narrow it down, and maybe even trigger it with just the SDK, but here’s the stack trace which FlexBuilder helpfully pointed me to:

!SESSION 2009-03-20 13:30:39.171 ———————————————– eclipse.buildId=unknown java.version=1.5.0_11 java.vendor=Sun Microsystems Inc. BootLoader constants: OS=win32, ARCH=x86, WS=win32, NL=en_US Command-line arguments: -os win32 -ws win32 -arch x86 !ENTRY com.adobe.flexbuilder.project 4 43 2009-03-20 13:33:28.109 !MESSAGE Uncaught exception in compiler !STACK 0 java.lang.ClassCastException: macromedia.asc.parser.MetaDataEvaluator$KeyValuePair at macromedia.asc.parser.MetaDataEvaluator.evaluate( at macromedia.asc.parser.SetExpressionNode.evaluate( at macromedia.asc.parser.MetaDataEvaluator.evaluate( at macromedia.asc.parser.MemberExpressionNode.evaluate( at macromedia.asc.parser.MetaDataEvaluator.evaluate( at macromedia.asc.parser.VariableBindingNode.evaluate( at macromedia.asc.parser.MetaDataEvaluator.evaluate( at macromedia.asc.parser.ListNode.evaluate( at macromedia.asc.parser.MetaDataEvaluator.evaluate( at macromedia.asc.parser.VariableDefinitionNode.evaluate( at macromedia.asc.parser.MetaDataEvaluator.evaluate( at macromedia.asc.parser.ClassDefinitionNode.evaluate( at macromedia.asc.parser.MetaDataEvaluator.evaluate( at macromedia.asc.parser.StatementListNode.evaluate( at macromedia.asc.parser.MetaDataEvaluator.evaluate( at macromedia.asc.parser.ProgramNode.evaluate( at flex2.compiler.as3.Compiler.analyze3( at flex2.compiler.mxml.InterfaceCompiler.analyze3( at flex2.compiler.mxml.Compiler.analyze3( at flex2.compiler.API.analyze( at flex2.compiler.API.analyze( at flex2.compiler.API.batch2( at flex2.compiler.API.batch( at flex2.compiler.API.compile( at at at at at com.adobe.flexbuilder.multisdk.compiler.internal.ASApplicationBuilder$MyBuilder.mybuild( at at at at com.adobe.flexbuilder.project.compiler.internal.FlexProjectBuilder.buildItem(Unknown Source) at Source) at Source) at$ at at at at$ at at at at at at at

Iteration Zero

I read this post today about “iteration zero.”  It’s a pattern I recognize well.

In fact I’ve been going through a lot of iteration zeros lately.  I think the preliminary work, including team selection, tool choice and setup, development environment installation, build and test process, etc., can be a significant factor in the success of a project.  Part of the idea of adopting Agile is to help define those processes and types of tools, if not the actual tools used.

There seem to be certain brands of Agile that go hand in hand with certain tools, such as a Java shop with Tomcat + Spring + Hibernate + Junit + Hudson + Confluence + Jira or a  Ruby team with Rails + Macs + Basecamp + CruiseControl.RB + FireWatir, etc.  Since certain experts (often the authors of specific tools) are strong advocates for their methodology, it’s no surprise that users of specific tools adopt a similar stack and brand of Agile.

I think we’ve learned, if nothing else, that the term “Agile” and the definition of  an agile process is slippery enough; much to the chagrin, no doubt, of the Agile “thought leaders” — who are probably as responsible as anyone else for the difficulty in pinning it down.  Part of iteration 0 is pinning down agreed upon definitions, but the important part is being able to create your “Hello World” using the tools and process decided upon.

Of course, each time a team repeats the process, it gets easier.  That’s the secret of a good agile team.  But the reality is that most teams aren’t static, and not surprisingly, neither is the technology.  If it were, I think a lot of developers would be looking for different jobs.  It’s the challenge of learning something new that appeals to many of them, myself included.

But there comes also time when you just want to “get ‘er done”, and that’s when the iteration zero becomes a pain.  Usually that’s a good sign, because it means you’ve got a project you’re interested in working on.  I said earlier that a good iteration zero was one contributing factor to success.  I think the other major factor is interest in the project.  Developer interest doesn’t always transfer to customer interest, but it’s a pretty good substitute.

So what am I most interested in?  I’m interested in the interation zero.  In finding the tools and process — for testing yes, but also for development, deployment, and project management — that best contribute to the success of projects.  Of course, that’s different for everyone, but there should be commonality, and much like Agile, I think there are good and bad patterns to be aware of.

My idea of a “QA Site” (a term I’ve never been completely satisfied with, but have stuck with) is to make iteration zero as painless as possible.  To reduce the ramp up time, and to build out the necessary infrastructure as quickly as possible, but keeping it lightweight (so it doesn’t become burdensome) and flexible (so it doesn’t bias to one set of tool preferences.)  To recongnize that there are both principles and preferences, and to not confuse the two.  I realized that I’m in the business of “Iteration Zero”.


ColdScripts is a good site with links, summaries, and rating for a bunch of PHP applications and frameworks. It has a few discrepancies and outdated listings, like WACT and Ambivalence for MVC frameworks (but, hey I still use Ambivalence for  That’s how I found it, after migrating Klamathsytems off of Westhost to share the same Linode as One Shore.)  The ratings looked fairly accurate and up to date however

Example categories include:

  • frameworks
  • blogs
  • cms
  • ecommerce
  • forums
  • galleries
  • wikis

A surprising absense was Magento from CMS

Give it away

I originally posted this as a comment under the 37 signals blog post How did the web lose faith in charging for stuff.

If you don’t have something of real worth, or aren’t willing to work hard to get your customers, or don’t have the skill to compete head to head in the market, then by all means charge for your software. There are always a few suckers out there to think Y costs more than X so it must be better. Your hobby project will probably net you a few dollars to buy candy after school or if you save it up, some little gizmo that has a light that blinks when you push a button.

Look at who’s talking here — 37 signals exists only because of free stuff. I can count the number of paying users they have on one hand (but only because I have six fingers on my right hand.) All of their money comes from consulting, because they invented two things and gave them away for free — rounded corners, and Rails. It is their reputation from giving away stuff that gets them clients.

Google controls advertising, so if you have viewers, you have value to Google. TV existed solely on advertising, and radio, magazines, and newspapers before it. So can the internet. The token subscriptions for print until recently only ever covered shipping. There is no shipping cost on the internet. If your application is 2 gigabytes, by all means charge a 4 cent download fee. But your competitors are going to laugh at you and pay 4 cents to everyone on the planet (that’s only $240 million) for the advertizing.

Free software *IS* advertising. And it’s better than a flash banner video clip with a wack-a-mole celebrity endorsement for herbal viagra delivered to your front door every morning by a spokesmodel. Because only people interested in your product will download it for free. And if they like it, they’ll tell all their friends. Or keep it secret and build a business model to monetize what you’re giving away. But then they’ll probably hire you as a consultant to improve it.

Project Foxfish

The last few days I took a diversion from my normal work to help out a worthy cause — my wife!  But she’s helping out an even worthier cause — OSSO.

OSSO is the Orphanage Support Services Organization, a nonprofit group that sends volunteers, supplies, and money to help orphans in Ecuador.  Before we were married, Kelsey spend 9 months in Cuenca, Ecuador volunteering for OSSO.  Shortly after we were married, we returned to Ecuador and lived there for 6 months.  While I did some of the early work on One Shore and came up with the QA Site idea, she did more volunteer work at some of the orphanages —  working with, though not a part of, OSSO.

Kelsey got pregnant, we returned to the USA, I took a job, she had the baby, and then I got laid off.   I’m doing freelance web development and test consulting now.  We thought about returning to Ecuador, but decided to stay in Seattle through the summer so I could work on my business here.

In the meanwhile, however, OSSO is in desperate need of both volunteers and donations.  We gave what we could, but we’re strapped too.  My wife came up with an idea, though.

She likes to make crafts (but doesn’t like to call them crafts.)  So far she’s made dozens of little blankets, toys, baby clothes, etc.  Mostly she gives them away to friends.  But the other day, she decided to open a shop on a website for crafts called Etsy at and donate all the proceeds of her sales to OSSO.  I helped her create a website for her project at

I installed Silverstripe CMS for her to be able to edit the site herself, including the gallery and ecommerce modules, as well as the hidemailto module I blogged about previously.  We chose a nice theme by Nivanka Fonseka at  I spend some time digging into the internals of the ecommerce module to try and set up inventory management and shipping calculations, as well as fixing (hack) some layout issues with the sunflower theme and ecommerce.  I’ll be diving into the ecommerce module more, fixing some bugs, and possibly working with the Silverstripe team on their Google Summer of Code project.

Anyway, that’s what I’ve been doing the past few days.  Kelsey’s been finishing up a bunch of crafts, and putting them on Etsy and setting up her website. We’re just about ready to launch.

If you know someone who likes crafts check out her site and Etsy shop.  Even if you don’t, but would like to help orphans, check out OSSO.  They could really use help, especially volunteers.  And I can attest that Cuenca is quite nice this time of year — 70 degrees and sunny pretty much year round.

SVN repository browsers


License: open source (custom)

I’ve installed ViewVC, and while it can be tricky to integrate, standalone it’s a snap.  It’s not the best viewer out there though, and my main complaint is that it seems a bit slow.


Creator: Atlassian
Java servlet
License: commercial

Fisheye is nice, but I never really liked it that much.  I’d give it a slight edge over ViewVC, but not to pay for.   Jira integration is a major selling point.


Creator: Tim Armes, Sebastion Petters, Dirk Thomas
Platform: PHP
License: open source

WebSVN is what Tigris (the makers of SVN) use to browse their repository.  It’s fairly basic, and being PHP, it’s fairly easy to install, except on Windows you’re bound to run into problems with the PHP subversion extension.


Creator: Horde
License: open source

Chora is part of the Horde project.  I’ve used IMP (the Horde webmail client) quite a bit.  Horde apps aren’t that pretty, but they’re fairly mature.  Chora, like ViewVC, takes a bit of clicking to get to the code, since it’s not too ajaxy.  But that can be a good thing.


Creator: BountySource
Ruby on Rails
License: recently open sourced

Yet another Rails SVN Browser (I wonder if the main functionality ships with webrick?).  I’m not knocking Bounty, it seems fairly nice.  I only tried their demo, and it seems quite responsive and simply laid out.   The more I poke around, the more I like it.  But I haven’t tried installing it.


Creator: ActiveReload
Platform: Ruby on Rails
License: recently open sourced

Warehouse gets good reviews for presentation, but I haven’t seen it.  I tried the demo, but got an error once and then an empty repository.  I wasn’t that impressed with the screenshots, though all I can say is the UI scores high on Web 2.0 (and I don’t even think it has rounded corners.)


Platform: Java servlet
License: open source (BSD)

I haven’t really used Sventon, but one feature I liked in the screenshots is the source view that shows the author next to each line.  It also gets high marks from users.  Probably one of the more featureful repository browsers, and looks like it is the most like browsing the repository manually.  Transition from a command line user is clear, like with big buttons for “blame” and “log” and other commands the svn expert is familiar with.  This will be on my list to give a second look.


Creator: Edgewall
License: open source

Trac has a repository viewer, but is primarily a wiki-based project management tool.  Or rather a wiki that can be used for basic project management, that includes repository viewing.  It also includes a simple issue tracker.  While I don’t use the issue tracker, it is a servicible, though basic wiki, and a decent code browser.  The strength lies in the combination of tools.  It’s a favorite of mine for a “design” centered wiki because you can tie design documents to the code.  If  Trac could integrate with more powerful defect and task management, I’d be even happier with it.


Platform: Ruby on Rails
License: open source (GPL)

I really like the Redmine repository view.  Much like Trac, it it more than a repository browser, though.  I think Redmine has one-upped Trac with a cleaner interface and more features. Redmine beats Trac for repository browsing, but I like Trac’s version comparison better.  Redmine is probably my favorite current viewer, though bsSvnViewer and Sventon look like contenders, though they lack project and issue management (which might be better), especially if I can integrate them with Bugzilla and scmbug and a QA Site dashboard.


Creator: Johan Sørensen
Platform: Ruby on Rails.
License: open source

Collaboa is another compsite tool that readily acknowledges their influence by Trac.  It looks more like Trac than Redmine, which is a plus in my book.  I didn’t see a demo, but one thing caught my eye, a future goal for supporting  “Continuous integration of tests/builds”.  I might want to get involved.

Of course the three tools I use most to browse subversion are: 1) the svn command line, 2) TortoiseSVN, and 3) the Eclipse Subversive plugin, but these really aren’t any good at distributed collaboration or integration with defect tracking or requirements management tools.  And they’re not web-based.