Unfortunately, I was able to find no information on recovery via binary files.  Most of the information assumes that you’ll be recovering from archived dumpfiles.  The Subversion book from Redbean has a section on migrating Subversion repositories, which is also useful for back-up.  It doesn’t mention binary recovery, though.

Fortunately, it turned out to be quite easy.

Step 1 – Copy the Repository Files

After mounting the old hard drive as read-only in the new server, I copied the repository files (which were in /srv/svn in my case) from the old drive to the new:

sudo cp -r -p /mnt/olddisk/srv/svn/* /srv/svn

The -r tells the copy operation to recurse through the directories, while the -p option preserves the permissions, ownership and timestamps.

Step 2 – Copy “Location” Entries from dav_svn.conf

Because our Subversion is set up for access over HTTP (to avoid problems with firewalls at remote locations), I had to make sure that /etc/apache2/mods-available/dav_svn.conf was properly configured on the new server.  First, I uncommented the “<Location /svn>” block and modified the authentication requirements.  I made sure that SVNParentPath was set to /srv/svn, since that’s where I put my repository files.

I then opened /mnt/olddisk/etc/apache2/mods-available/dav_svn.conf and copied the other “location” blocks from that to /etc/apache2/mods-available/dav_svn.conf. We have multiple repositories (one per client), so each repository requires its own entry. You could have a single entry and allow users to see what all of your repositories are, but that’s less secure.

Step 3 – Recreate Users

Finally, you have to recreate your users. We keep good records, so that was fairly easy.  Once this step was completed, we were able to access our old repositories immediately.

I hope that this post saves someone the research that I spent on trying to figure this out.

I’m using QuickBooks Pro 2010.  I have a love/hate relationship with this software.  I love that it does almost everything that I need.  I hate that Intuit plays vendor lock-in in order to sell all sorts of other products and services.  Unfortunately, I haven’t seen any open source software that does what I want, and I’m not going to trust my business’s critical financial data to a SaaS.

All of the information that I found was geared toward issuing a full refund of an overpayment or issuing a credit memo (which creates a new, separate credit) and then refunding that.

After lots of Googling, I finally put the answer together:

1. Click on Customers -> Create Credit Memos/Refunds
2. On the Credit Memos window, click on the “Use Credit to” button on the toolbar and select “Give refund” (really obvious, right?)
3. You’ll be prompted with “This credit memo has no amount to be refunded.  Would you like to create a refund check that is not directly linked to this credit memo?”  Answer “Yes”.
4. Enter the name of the customer/job in “Pay to the Order of”.
5. Enter the memo, select “to be printed” or enter the check number, etc.
6. In the account field, enter “Accounts Receivable”.  This is critical.  If you don’t, it won’t be applied to the customer’s balance, and they’ll still have an outstanding credit.
7. Finish your other options and save the check.

I don’t know why Intuit hid the ability to issue refunds not connected to credit memos (hate) or why I couldn’t find any useful information about how to do this in the help (hate hate), but at least I can do it.  I hope that this post saves someone else the time and frustration of trying to figure it out.

In my experience, most people either overestimate or underestimate the difficulty of building software, usually by a significant degree.  People who overestimate tend to be those who actually use the software to do their work.  People who underestimate it tend to be managers and other decision makers who have to pay the bill.  No surprise there.

As a consultant who routinely deals with new clients, a big part of my job is to educate them on the level of effort required to develop a piece of software.  It’s not uncommon for the first consultation to end in disappointment when I inform the client that their new YouTube killer cannot be built for $2,000.

While the disparity between budget and desired outcome is frequently a factor, I think that there are others.

First, because people routinely pay small sums of money for software that has a huge number of features, they estimate the labor that went into the software according to the price they paid.  While they may know that other people also bought that software, the actual scale of sales isn’t evident so they can’t extrapolate the cost of production and thus the amount of labor that went into it.

Star Trek syndrome may also be at play.  By this, I mean the idea that you just tell the computer what you want, and it does that.  Technically, that’s the way it works, but on a much, much lower level than most people realize.  Even with modern GUI, database, etc. toolkits, you still have to give it instructions like “Put the button at coordinates 10, 30.  When it’s clicked, check the first checkbox.  If that’s checked, set this variable.  Check the next checkbox.  If that’s checked, set the other variable.  Open the file.  If that failed, check the error code.  Now display this error message.  If it didn’t fail …”.

Finally, a lot of people never really sit down and document the work processes that they execute every day, trying to account for all of the inconsistencies, errors that may occur, exceptions, etc.  When you’re writing software, particularly custom software for a business, you have to do exactly that.  It’s complex.  There are all kinds of moving parts, and exceptions and inconsistencies abound.  That analysis and documentation process alone can take days, particularly when you have to pull information out of the brains of multiple people who aren’t used to thinking in terms of process.  It’s like asking a baseball pitcher to tell you the exact motions he uses to throw a curve ball.

The next time a client tells me that something can be done in a few hours, I think that I’m going to point them to Hilton Lipschitz’s blog post.

So far it’s nice, but one thing that was missing was the icons on the desktop.  I read through a few forum postings, then finally found this blog post by Leonid Mamchenkov.  It mentions using gnome-tweak-tool.

I installed gnome-tweak-tool using yum, then ran it:

$ sudo yum install gnome-tweak-tool
$ gnome-tweak-tool

The application came up, and I selected “File Manager” and noticed that the “Have file manager handle the desktop” option was off.  I turned it on, logged out, logged back in, and voilà, my icons reappeared.

I don’t know why that page was crashing, and I didn’t really care.  I just wanted to close that tab so that it would stop crashing my sessions.

After some research, I discovered that Safari stores the open tabs in ~/Library/Safari/LastSession.plist. Deleting this file got rid of the existing session, so I was able to start Safari, create a new session, and not have Safari crash on me.

There’s probably a way to edit the plist and just remove the tab that was crashing, but that would involve using a plist editor, and I didn’t want to save the session enough to justify going to that trouble.

This isn’t a stupid question, but his post runs off the rails in a couple of areas.

First, he asks “If you are building a company, would you prefer one standout person over one hundred pretty good people?”  I see two problems with this question.  The first is that it poses a false dichotomy; your choices aren’t limited to one excellent person or one hundred pretty good people.  While your payroll budget is always finite, in most professions you can hire someone truly outstanding for a fairly small multiple of the cost of getting someone who is pretty good.  This is particularly true in technology, where you don’t see huge ranges in sallaries.  Software engineers might make between $30,000 and $250,000 a year depending on experience, skillset and security clearances, but that’s a far cry from the differences in industries like professional sports, music and movies.

The second problem with it is that the answer really depends on the business that you’re in.  If you’re churning out widgets in a factory, then numbers are important.  If you’re designing software, then too many cooks spoil the soup.

Mr. Taylor then goes on to ask “If you were launching a technology or developing a product, would you rather have five great engineers rather than 1,000 average engineers?”  Yes, I would absolutely rather have 5 great engineers.  Great people produce great work, while average people produce average work.  Furthermore, a committee made up of 1,000 average people is likely to produce a monstrosity.  When it comes to software, every programmer contributes at the creative level because writing even the most mundane piece of code is a creative act.  It’s hard to explain to people who don’t write software, but think of it this way: software is a set of instructions, like a recipe.  Can you imagine a huge wedding dinner where 1,000 people wrote the recipes, even if that recipe is only for a sauce or combination of seasonings?

He starts to reveal his root question with the following: “[are] we are prepared to shower millions of dollars (maybe tens of millions) on a small number of superstars rather than a well-assembled team that may not dazzle with individual brilliance, but overwhelms with collective capability?”  Continuing, he brings up the example of the Spanish pro soccer team, FC Barcelona, which has been a highly successful team, mentioning an essay in The Economist that emphasizes Barcelona’s coaching emphasis on team play rather than individual superstars.

There are plenty of examples in history of well organized, cohesive units who overcome opponents staffed with a few superstars, but lacking in team coordination.  Barcelona is just one good example of that.  However, I don’t think that anyone would argue that Barcelona lacks superstar players.  In fact, according to the essay quoted, Lionel Messi is a team member and is widely considered one of the best players in the world.  It would seem that one can have both team strength and outstanding players.

In the final paragraph, Mr. Taylor concludes, “I’m certainly not suggesting that leaders who are growing companies or building teams should settle for mediocrity. But I am suggesting that there is more to long-term performance than the excellence of your individual players.”  The first sentence seems to me to be at odds with questions asked and statements made earlier in the post.  Perhaps Mr. Taylor was using hyperbole to make a point?  In contrast, I would agree wholeheartedly with the second sentence.  In its light, his complaints make more sense; it may be arguable that some firms have gone overboard in chasing talent as he argues.  However, I don’t think that his arguments illustrate that point well.

Also, when discussing the amount of money spent, it’s important to remember that the software industry has been growing and competition for great technical people has been heating up.  Google really lit the fire with their hiring drive and a publicly declared focus on hiring only the top people.  It seems like an obvious thing to do, but in reality too many companies had hired top people as software architects or designers and then filled the “grunt work” positions with people who were inexperienced, of average (or below average) ability or both.  Now the industry is playing follow the leader and chasing talent.

That doesn’t mean that all of them are blindly throwing money at “rockstars” and ignoring everything else.  In fact, if you hang around in the DC tech start-up scene, you’ll hear a lot of discussions about building great teams, reevaluating practices, and using techniques that facilitate communication and collaboration (both of which are essential to working in a team).  LivingSocial is a good example of this.  I’m acquainted with a some of those guys via the DC RUG and they’ve all struck me as bright, focused programmers who enjoy collaborating with other smart people.  Not a single one of them has come off as a prima donna.  I think that their current success is also indicative of their ability to work as a team.

In the end, I think that you need the best people that you can find to do the work, and a management staff who can create and manage a team effectively.  Note that my definition of “best” includes being able to work effectively as a member of a team.  Prima donnas and lone wolves need not apply.

Google lead me to this somewhat vague page on J!BT Websites’ site.  Here’s the short version. Every module is assigned to a position.  This position controls where on the page the module appears.  Check your module to see what position it’s using.

Open up your template’s index.php and look for a line like: <jdoc:include type="modules" name="right" style="xhtml" />. You’re looking for one where the “name” attribute contains the position that your module is using. Is it missing the “style” attribute? If so, you need to add it.

wrong: <jdoc:include type="modules" name="right" />

right: <jdoc:include type="modules" name="right" style="xhtml" />

Not having the style attribute causes Joomla! to not add a number of important elements, like the module title.

I went a’Googling and found a page dedicated to the error on the MozillaZine site. That page lists a number of Firefox add-ons that are known to cause this error. Imagine my surprise to see Extended Cookie Manager on the list. I’ve been using that add-on for years, since version 1.x, to allow me more finely grained control over the cookies that sites set. It didn’t cause this problem in the past, so I wasn’t sure that it was responsible now.

I disabled Extended Cookie Manager and tried Google Maps and Meetup.com. Both appeared quickly, with no delays when navigating around the site. I had been forced to use other browsers for months to avoid such problems, and now they were back to normal. I had my culprit.

A quick visit to the Add-ons directory yielded Cookie Monster by Tony Schilling. It does everything that Extended Cookie Manager did and more, without slowing down Firefox.

It’s too bad that I had to switch, but Extended Cookie Manager hadn’t been updated in over two years, so its future looks dim. Check out Cookie Monster and enjoy surfing without errors.

Searching on this error brought me to a very useful blog post by a person named Schleichermann that described the same error, although he was using TortoiseSVN rather than RapidSVN.  The error was being caused by the Windows Indexing Service, which is locking some of the files in the .svn directory.  You can exclude folders and files from indexing, which is the solution here.

Options:

1. Turn off indexing for the top level folder.  Pro: you won’t have to remember to add .svn folders in any subfolders to the list of items to be excluded from indexing.  Con: it won’t index source files in those folders either, which will slow down searches in those folders.
2. Turn off indexing for only .svn folders.  Each must be done manually.  Pro: The indexing service will still index the source files, making searches faster.  Con: every time you add a subfolder to the project, you’ll have to remember to add its .svn folder to the list of folders excluded by the indexing service.

I tend to favor option 1, as I rarely use the Windows search on source code.  Instead I grep it.

In order to make it easier for him to check his changes on the fly, I helped him install and configure IIS under Windows 7, then set up the application to run locally.  I had compiled everything on my system and pushed it over to him.  Since he would only be making changes to aspx files and CSS, there would be no need to recompile.

Once we had IIS configured, we copied the files into the wwwroot directory and ran the application.  Immediately, we got an exception:

Exception Details: System.Security.SecurityException: Request for the permission of type ‘System.Web.AspNetHostingPermission, System, Version=3.5.0.0, Culture=neutral, PublicKeyToken=c3d6b267409e098c′ failed.

There was clearly a problem with the permissions, but exactly what type of problem wasn’t clear.  We validated the permissions on wwwroot and confirmed the account under which IIS was running.

After reading through a number of forum threads and blog entries, I ran into this post from the French IIS Support Team.  It pointed out what should have been obvious.  The code was compiled on another system, and so would be treated as suspect by the system until steps were taken to declare it safe.

If you’re used to native programming from the pre .NET days, this may seem a little weird.  However, Microsoft introduced many more security options with .NET.  Programs can be granted varying levels of privilege, from not running at all to having complete access to the system.  The system defaults tend to be cautious in an effort to protect users from accidentally running something that they shouldn’t.  The compiled code in this case hadn’t been installed or signed by a recognized authority, so it was prevented from running.

The solution was to check the DLL’s properties and click the “unblock” button at the bottom of the General tab.  Once we did that, the security exception disappeared and the application ran smoothly.