Either You Succeed or Explain

Eclipse on Mac: Use Magic Mouse/Trackpad back and forward gestures

The new Apple Magic Mouse is a controversial piece of hardware.  Most people either really hate it or adore it. Personally, I think it is probably the best mouse I've ever used. There's a lot of criticism regarding the low profile of the mouse, suggesting it is not ergonomic. From my experience, I don't have any more wrist pains since I started using it. Ergonomics aside, the highlight of the mouse is the upper multi-touch surface with the enabled gestures.

I use the back and forward gestures a lot. Especially when browsing the web. Swipe two fingers to the left and go back. Swipe right to go forward. It is very easy to get used to it. It works in web browsers, it works in Finder windows and native applications are adding support as well. However, it doesn't work in Eclipse. I want to swipe back and forward when browsing code. Back, go to previous location. Forward, return to the next location.

I incidentally found a solution for that. There are many programs on the market that augment the Magic Mouse behavior. The reason for their existence is because Apple provides very limited gesture functionality. Other than back/forward and scroll, there's simply no support for other functions, not even Exposé or Spaces which were supported in the previous Mighty Mouse and are supported on the multi-touch trackpads. The most popular tools are MagicPrefs and BetterTouchTool (both free) but there are many others, free and commercial. Personally, I use MagicDriver, which is commercial (free while in beta). The reason I prefer it is because it has much lower CPU utilization, which was an issue for me in MagicPrefs.

MagicDriver replaces the back/forward gestures with their keyboard equivalent: ⌘+[ and ⌘+]. These shortcuts are commonly used in OS X. Eclipse, by default, also uses these keyboard shortcuts to navigate back and forward. It just works. MagicPrefs and BetterTouchTool will require some customization: you can define the two finger swipe left and right to fire these keyboard shortcuts rather than use the default back/forward functionality.

If you use a newer MacBook with a multi-touch trackpad or a Magic Trackpad, you can achieve the same functionality by using BetterTouchTool. AFAIK MagicPrefs does not support it and the current version of MagicDriver doesn't support it either.  BetterTouchTool also has the ability to define gestures per application, so you can customize the behavior specifically for Eclipse and leave it as is for the rest of the applications.

If you are new to these tools, I should warn you: defining too many gestures doesn't work very well. There are tons of options and it is very easy to get carried away and use as much as you can. However, there's probably a reason why Apple did not include support for all those gestures in the first place. It is very easy to "miss-fire" and perform gestures by accident. You don't always pay close attention to the number of fingers you have on the surface, so mistakes are very common. I just use a 3-finger click for expose. Don't be greedy and it will work just fine.

Finally, if you want proper native support for back/forward gestures in Eclipse, you can vote for this bug.

People are willing to work for less

Tweet

Last week, I wrote a post about my experience with oDesk. The purpose of the post was to share my experience, so other entrepreneurs can learn from it, and provide feedback to oDesk and contractors in general. While some people found the value, I got some pretty angry comments on reddit. It is worth clearing up some of the issues discussed there.

First, it should be clear that I am not bragging about my actions or how I managed to exploit a person. As I wrote in my post, from my communications with Max (the service provider), I believe he was happy with the deal and will work with me again at similar conditions. The job itself didn't take 2 full days as my post implied. It took about 6-7 hours of work over the course of two days.

We can all agree that $37 for 7 hours of work is low salary. QA people at western countries are probably deeply shocked by these numbers. Well, it's not just their problem, it's everybody's problem and it is called Globalization. Look at the crumbling European economy. Jobs are migrating east as unemployment in the west grows. You can fight it, you can resist it, but you cannot deny it. If geographic location is not an issue, there are plenty of people willing to work for less. In some cases, it's not enough for them to get by, so they will remain poor working for low wages.

The question is: should you hire them at such low wages?

If you have a job and they are willing to do it for a low price, why not?   It might seem obvious to some, but it is not a simple question. First, you need to stay on the moral high ground. I would not hire a 12 years old QA person, not matter how low the price will be. Sweatshops are another examples of how to exploit workers and nobody should support them (chances are, you paid for products made is such facilities, but let's not go there). A notable example is the recent wave of suicides at the Foxconn factory in China, the same factory which makes iPhones and iPods. It created a negative backlash, eventually reaching up to the CEO level and resulting in salary raises.

So, we can all agree: if you have some decency, you will not hire any person for any price, even though you can. But where do you draw the line?

Personally, I would like to have a good QA person which I can trust and I'm willing to pay him a reasonable hourly rate. Reasonable means reasonable according to his local standards. But it is not that simple because:

• I really don't know if the person I'm hiring is any good at what he does. If he has some feedback, it helps mitigate this risk.
• I don't trust the person I'm working with. If he says he was working for 5 hours, was he really working for 5 hours?
• I need someone productive. Even if he was working for 5 hours, how is his productivity?  If I estimated the job will take 5 hours and it took him 20, that's not very helpful for me.

From all those reasons, it is easier for me to start with smaller tasks that have a well defined budget. Low budget, because I might not get the results I was hoping for, so I want to minimize the risk. I hire people with no (or minimal) feedback and credentials. It is much like using an intern, a student or a new college graduate. Once we get comfortable and build trust, we can move forward and change the terms. Perhaps, once they get good feedback they will decide to move on to a more lucrative offer which they can now win. This already happened to me with a previous QA person and I'm really happy I could help him get a better job and move forward.

That's how the employment market works. Experience and credentials will get you a better salary. You start from the bottom and you work your way up. Note that I didn't get any offers from people with well established feedback. Perhaps because they can find better offers.

Final note to the potential employer: you can continue using the services of the cheaper guys, but you will still have all the risks. Or, you may establish good working relations with a few good providers, minimize the risks and get better results with less "hand holding" from your side.

My experience with outsourcing QA using oDesk

Tweet

About a month ago I decided to host Eclipse distribution downloads on the nWire web site. Alongside with the standard distributions, you will find distributions that already include nWire pre-installed. Packaging and repackaging was a very simple task and, on the face of it, everything was in order. Nevertheless, I decided to do the right thing and test that everything works.

The challenge is to test on as many platforms as possible: Windows XP, Vista, 7 and Linux, on 32-bit and 64-bit. It was narrowed down to 24 tests of unzipping and running Eclipse. Our resident QA engineer was preoccupied with other tasks and didn't have all the platforms installed, so I decided to outsource this task.

Choosing a Service and Signing-up

There are many sites that offer freelancers for hire. I heard of oDesk and Elance and they seem to be the biggest sites to date. Other sites worth mentioning are Guru.com and vWorker (aka rent-a-coder). Elance requires a $10 registration fee, so I decided to try oDesk. Plain and simple. On the other hand, unlike Elance oDesk does not work with PayPal. Since I use PayPal to sell my software, paying with PayPal is very cost effective to me: it saves the very expensive round trip of converting USDs to my local currency and vice versa. In the long run, I would prefer paying with PayPal.

The sign-up process is quick in oDesk, but, after you sign up, you need to verify your credit card. oDesk will charge you with a random fee (still hope to get it back as promised) and you need to enter this fee as a verification. You cannot hire before you verify, at least for payment up-front contracts. This turned out to be an annoyance, since I wanted to get this job out the door as quickly as possible. I had to wait for the charges to appear in my account and than phone the credit card company and ask for them. At hindsight, I'd rather pay $10 on PayPal...

oDesk has a desktop client. You don't need it if you're hiring. On the Mac, it looks really out of place, so I recommend not installing it if you don't have to. They also have an iPhone app which I haven't tried.

Posting the Job

Posting the job was easy. The interface is simple and intuitive, although somewhat outdated. I was very careful to specify the exact requirements and I think I did OK (you can read it yourself). I estimated how much time it will take, checked for common rates and decided to make it a fixed-price contract at $30. I like knowing in advance how much it's going to cost, especially when I don't have any experience (read: trust) with the person I'm working with.

Sorting Applications

I got about 20 job applications. The job was posted on July 13th. Here's the distribution of applicants per day: July 13 - 5, the next day 5 as well, then 3, 1, 3, 1 and on July 19 - 2. The last two were spam. Completely spam. They both offered to do it for about $250 which was absurd. Other than those two, all offers were bone-fide.

Sorting the applicants was frustrating. The list kept growing and it was hard to keep track of the conversation with each applicant. I was really missing a way to write notes for each applicant. The applicants list UI is poorly executed and very annoying, mostly because the sort order keeps changing and opening new windows is inconsistent.

Candidates came from India, Western-Europe and Russia, Philippines and South America. I received many applicants from India, over 50%. None of them read the requirements properly. It was required to have access to all the testing platforms. I had to ask each applicant individually whether they have it and all said no. It was really a waste of time.

Finally, I managed to find some applicants that had the platforms. I chose Max. He had relevant experience, but not on oDesk. He was new, like me, and didn't have any feedback. We scheduled to talk over Skype. He's from Ukraine, so we are at the same timezone. I sent him a full specification of the tests before the conversation. We talked, I was under the impression that he understands the job and we moved on.

His offer was $33. The majority of applicants added 10% to the price. Some did the opposite. You might want to subtract 10% from what you think is fair, knowing offers will add 10% anyway.

Execution

Since it was a fixed-budget job, I didn't care about getting a work log. I just wanted the job done as quickly as possible. Max was quick to execute and provided organized screenshots of all relevant scenarios as a proof that it works. We used Dropbox (here an affiliated link) to share the files.

One of the tests failed because the distribution was corrupted. Max also had some issues with the 64 bit platforms. I had to step in and help in solving it, but he was proficient: I just provided the links to the solutions and he did the rest. Overall, it took about 2 days because of the 64 bit issue. Once the tests were complete and the corrupted file replaced, I now can safely say that everything works.

Eventually, I was happy with the results. I added 10% to the agreed price, which totaled at $37. I think it was a fair deal.

Conclusion and Tips

I got what I came for, and I got it at a good price IMHO. I'm happy. I think the key was good definition of the task which led to choosing the right person.Choosing a person with no oDesk feedback was not a problem in this case.

I will use such a service again in the future. Will I use oDesk again?  Maybe. Not having PayPal as an option could be a deal breaker when the stakes become higher. I didn't like the interface and it is infested by spam to some extent.

While looking into other sites, I found that some sites are better for specific needs. For example, some sites have more cheap QA people while other have more talented marketing people. If you have a specific need, I suggest running a search for your target employees on each site and comparing results.

If you are going to post a job, make sure to highlight the requirements. Prepare a canned list of questions verifying the applicant fits the requirements (even though you specified them in the description) and send them to each applicant without even reading the application or the applicant details. You see, many freelancers just send canned applications without even reading your requirements. When you send questions, they will need to answer them. If they don't answer, forget about them. If they answer that they don't fit the profile: thanks, but no thanks. Finally, you will get a short-list of people who actually read your questions and understand what you need.

Good luck.

Serving static web sites from Amazon CloudFront

Last week, Amazon made a very small technical addition to the CloudFront service that has the potential of rocking the hosting world.

I haven't written anything on cloud computing yet. I'm no expert, but I do use several services, so there's no reason why I shouldn't. So far, I have the nWire update site hosted on Amazon S3 + CloudFront and the nWire Eclipse distributions hosted on Rackspace CloudFiles. Rackspace is cheaper, but not suited for hosting Eclipse update sites because it is limited to a flat file list with no support for folders. But that's not why I'm writing this post (and possibly a topic for other good posts).

Back to the main issue: Amazon just enabled having a default root object in your CloudFront distribution. Essentially, this means you can now host complete static web sites on Amazon CloudFront with no need for any proxy or web server what so ever. Up until now, you could host all your static content in CloudFront. For example, if your main page was index.html, visitors asking for http://www.example.com/index.html would get what they need. But when a visitor makes a request for http://www.example.com/ the server would simply answer which a response that looks like gibberish. Not anymore. Now, you can tell the server to direct the visitors to index.html, which means you have a web site.

CloudFront is a very cost affective and highly available combination. Such a static site will probably withstand traffic spikes better than any other solution you can think of. It is served from the CDN edge locations, so it is also very fast. The cost is really dirt cheap compared to quality hosting solutions. More importantly: you pay for what you use by the amount of data and you don't pay for CPU time.

The downsides of such an approach is that you cannot use dynamically generated pages. Many of the informative sites I visit are mostly static. Especially company web sites. The dynamic parts are the blog and the forums. One can easily create a combined solution where the blog is served by a service like Posterous and the forums are hosted by a service like Zendesk. Forms can be hosted by services like FormStack and a simple mailing list signup can be embedded using ready-made JavaScript code from MailChimp.

If you really must have dynamic content you could also consider using AJAX to load it into the static pages (JQuery makes this a no brainer, but I digress). For example, if you have badges from services like TweetMeme or Reddit, this is exactly what will happen.

Another downside is updates. CloudFront takes time to propagate updates to the edge locations. Amazon claims it may take up to 24 hours. My experience shows it will take much less, but it is not immediate. Usually it will take a few hours.

At the moment, getting started is not trivial. Setting up the root object requires some hacking (see the CloudFront documentation). The CloudFront console still doesn't support setting this property. I've been told that the next release of CyberDuck will support setting this root object, which will make the job a whole lot easier if you use a Mac. I'm sure that Windows-based tools will add support as well.

Eclipse Plugin Installation and Windows User Access Control

I make Eclipse Plugins and I sell them to developers using Eclipse. Most of the visitors to my web site are not Eclipse experts. One of my key challenges is getting them to successfully install and try my software. It sounds simple, but Eclipse plugin installation is somewhat complex and error-prone process.

The safest way to install a plugin in Eclipse is by using the update manager as opposed to using the Dropins folder or manually copying files. It is safer because Eclipse performs more checks to ensure everything is in order. For example, it can check that all the dependencies are met.

Lately, I got several reports from users who failed to install my plugin using the update manager. After some research and with the assistance of some devoted users, I located a major issue: Installing Eclipse plugins on Windows Vista & 7 with User Access Control (UAC) on, fails if Eclipse is installed in the Program Files folder.

It is a very common scenario: many users are already running Windows 7, the UAC is on by default and the Programs Files folder is where most of the applications are installed, especially custom Eclipse distros that come with fancy Installers. It is a frustrating issue because the plugin installation appears to be complete: it will not issue any error message and, after restarting the platform, it will even show up properly in the Installation Details box. However, the plugins and features JARs are not copies, hence the plugin in not installed.

There are several possible work arounds for this issue: (starting from the easiest)

• Before the installation, start Eclipse with "Run as Administrator" option.
• Turn off the pesky UAC (for the installation or, better yet, altogether).
• Install Eclipse in a different folder. Eclipse doesn't have to be in the Program Files folder.
• Switch to Linux or Mac.

Clearly, the way Eclipse is handling it is bogus and should be fixed. This problem has been around since the days of Vista and Eclipse 3.3. I understand that installing a plugin in this case is an issue, so the least Eclipse should do is fail the process and not mislead the non-suspecting user. If you agree, you might want to vote on this bug (there are several bugs on this topic).

PS. If you're looking for more useful tips on installing Eclipse plugins, check out this article. It was written for Galileo but still applies to Helios.

PS2. I know, I wrote an entire article on Eclipse plugins installation and never once mentioned P2. So, there, I mentioned it.