Today is one of those days that I really love my job. I’m on a train headed towards New Orleans, LA for RubyConf happily catching up on my digital backlog of newsfeeds, podcasts, emails, Instapaper articles, emails, watched github repos, and all of the other digital detritus that builds up in my “defer” bin during the normal day-to-day of consulting. The backdrop to all of this is the countryside of the deep South rolling by at a steady pace on a sunny late fall day. I’m honestly thinking I might need to take a trip like this every few months just on general principles.
This year I’ve worked in a number of different environments ranging from a week at the bar at a Pacific beach resort (I *did* get up at 4am, so it wasn’t all palm trees and rum drinks) to my temporary summer office at an 18th Century Hacienda in Oaxaca, Mexico to my corporate office on the Court Square of bucolic Harrisonburg Virginia. I’ve worked onsite with clients in hip downtown office complexes, funky small-town lofts, and coffeeshops of all shapes and sizes. Despite the wide differences between all of these disparate locations, the only two prerequisites for me to be productive are somewhere to plug in to recharge the Macbook and an internet connection.
Power tends to be a given, despite the occasional need to hunt for where the outlet is hidden at a particular coffeeshop or the need to have the right adapter for international travel. The recent increases in battery life in newer laptops, particularly the 5-7 hours new MacBook Pro and Air models, means a solid amount of work can be done on a single charge. Electricity is no real impediment to the mobile developer.
Internet on the other hand, can be a huge problem. In Oaxaca, DSL was widely available, but it was pretty slow (512k usually gave me about 50k download) with huge latency. In Harrisonburg, the local cable company won’t wire my downtown office which gives us 2 bonded 768k DSL lines as our best inexpensive option. And of course if you’re truly mobile, you’ll need some sort of WWAN cellular solution.
I’m no AT&T fanboy — I used Suncomm back in the day, kept them and my Treo when AT&T bought them because it was easy, and then was pumped that I could upgrade to an iPhone when it came out. Of course as we all know, AT&T’s network is pretty spotty as far as coverage which makes it an awful solution for traveling and expecting an WWAN internet connection to work. To add insult to injury, making my iPhone 3gs into a hotspot costs $60 for 2gb of data and more importantly TAKES ME OFF MY UNLIMITED PLAN FOR GOOD. Not cool AT&T.
I got a Verizon MiFi this week for the RubyConf trip. So far, I’m impressed. Setup was generally easy though you have to install a piece of software (VZAccess) and reboot your computer so you can basically press the “activate” button on the modem, which seems like a lot of extra work, especially since the software is then no longer needed. Once it was setup, I reconfigured it to use WPA2-AES, added the MAC addresses of my phone, iPad, and Mac for a little extra security, and started happily surfing the web from all three devices.
It’s been particularly interesting to watch my iPhone connectivity along with the MiFi — my definitely non-scitentific observations of the connectivity of both through VA, NC, SC, GA, and AL is that both connect *great* in the city and AT&T basically dies once there are trees and other outdoors. I would *definitely* move to a Verizon-backed iPhone based on how much coverage and speed I get.
While I was writing this, I downloaded the 11mb release of jRuby over the MiFi traveling somewhere in Georgia in 130s with an average spped of around 88k with spikes above 130k. Wow. I could tell you where, but the wifi is off on my iPhone and it can’t download the maps since there’s no AT&T service, though the GPS signal is fine. In summary, I’m feeling pretty lucky to be able to work like this — where I want, when I want — even if that place is on a train, a rural Virginia town, or 11,000 feet up a mountain in Mexico. Or the Big Easy for RubyConf.

1. Create your Fluid App as normal
2. Go to Preferences > Advanced
3. Add *launchpad.37signals.com* to the list of allowed sites

For extra credit, you can add the related URL paths for any of the other 37signals webapps so you can use the menubar you get on their sites when you’ve authenticated with launchpad.

Basically, if you want to use the sexy new WP capabilities for automatically upgrading, directly uploading themes, etc you want to ensure that the process running WP can access the files. I had uploaded WP as a user, so my list files were owned by ashenfelter.com:ashenfelter.com while new files created by WP plugins (eg sitemap.xml)  were owned by www:www. Solution?

web/public >sudo chown -R www:www .

As dozens if not hundreds of posts on the WP forums/etc warn — please DO NOT try to solve this kind of problem by chmod 777 unless you like being hacked.

One the plus side, I have 3 lifetime accounts — yes, lifetime. When TextDrive, the original company, was bootstrapping themselves, they took the unusual step of offering several rounds of VC (venture capital) lifetime hosting accounts. Breakeven point compared to monthly hosting was about 16months as I remember so considering I’m on year 5(?) I think I’m pretty ahead of the game.

For those that care, I have the equivalent of two Shared Accelerator accounts and one Dedicated (M) Sized Accelerator. Since Joyent bought TextDrive, everything has migrated from FreeBSD to their Solaris based virtual machines (sort of like Xen on Linux). I finally migrated my FreeBSD boxes onto their Shared Solaris machines this month and restarted my long dormant WordPress blog.

Install on the Shared Accelerator was a breeze! The VirtualMin control panel has an install script for WordPress 2.8.5 which then could autoupdate itself to 2.9.x. The only problem is that WordPress on Shared Joyent Accelerators is so slow as to be unusable. Seems like there’s plenty in the forums about it. And by unusable I mean 5-15s (SECONDS!) to render a simple WP post. Admin side could take well over 30-60s. I installed WP-SuperCache and basically had no real improvement. #fail.

Fortunately I had the Dedicated Accelerator. Turns out WP actually runs pretty well on that machine. It’s dedicated space, so it *should* run faster, but my Shared FreeBSD server ran circles around the equivalent Solaris server as well. Chalk it up to shared hosting and it’s resource sharing versus dedicated resources for virtual machines.

Of course I had to manually install everything and then fight a bit with permissions to get everything working. Seems like PHP hasn’t changed *that* much in the years I’ve been avoiding it! But you’re looking at the final result — my personal ashenfelter.com site and my corporate transitionpoint.com site both running WordPress on a Solaris virtual machine.