For the last 6 months I've been acclimating to OS X on my MacBook Pro. Having a day job that necessitates having a modicum of Windows familiarity and a strong affinity for the most masochistic of Linux distributions (Slackware, for those in the know), I've found in Tiger, the holy grail of my various OS meanderings (albeit a rather battered and less then golden artifact).

Let me say it right now: OS X is not perfect.

Although the idea of having a single place to do open source development with the robustness of UNIX, and an end-user experience second to none, definitely appeals to me, I shy away from outright Mac-boy fanaticism that characterizes most Mac owners. The reasons are simple: there are some glaring deficiencies that irk me and interfere with my rather specific usage of my machine. Notice that I use the word "irk" here. They are truly minor annoyances; not enough for me to flee back Windows/Linux. More on these problems later.

Although Boot Camp has been available for some time now, I've been pleasantly surprised by my utter disinterest in dual booting into another OS. Almost everything I need is available for OS X, and if its not, there is nothing stopping you from finding a suitable alternative in the form of pre-compiled binaries or source code. I subscribe to half-empty-kitchen-sink philosophy so this state of affairs suits me just fine. For those who absolutely must install every visual gadget, this setup may seem less ideal.

There is an additional unexpected benefit in my shift: the Mac really does allow you to focus on the task at hand without all the visual distractions that are common in Windows. In other words, it lets you focus on work. Its not quite as spare as the command line, but pr0n just doesnt't look great in ASCII ;^p. For someone who had a definite preference for the command line, this came as quite a surprise. The uniform application of anti-aliased text as well as the ability to set the back light to the low radiation/stun only setting has noticeably improved my reading speed when using my Mac. Although the Windows equivalent Clear-Text is available for WinXP and newer, there is definitely a difference in their implementations.

A new user may be tempted to splurge on purchasing Mac software and although the base Mac installation is already feature rich, there are just a few necessities missing. Image manipulation, package management for open source software, and a productivity suite for manipulating Microsoft Office documents, are just a few of the shortcomings. Although, iWork addresses a few of these issues, I opted to use software that is available in the widest number of environments (Linux/Microsoft/OSX). The following list fills the software gap on the Mac without compromising anything in terms of quality or functionality:

• XCode

  Having come from a Microsoft Visual Studio environment, I was pleased by the availability of developer tools for OS X without the additional cost of having to purchase and IDE. Just poke around the installation to CDs to find this great collection of tools. This package is required if you plan on compiling your own programs for OS X.

• X11

  There are just some applications that do not have native Aqua support. This allows those applications to run side by side with your native OS X apps. Granted these apps do not make use of the Aqua UI elements, but it is a small compromise to run some great software.

• Parallels

  The environment of last resort. If you can't find a native OS X version and it can't be compiled from source, you will have to rely on a virtualized Windows environment. Surprisingly, things like passable tax software are difficult to find for OS X. Never fear; their is an alternative for these situations as well.

• Fink

  Portage on steriods. If you don't that open source app on your system, fear not. Fink allows for compiles directly from source with dependency tracking.

• Open Office

  I've used prior version 1.x of this suite and was always disappointed with the format mangling that inevitably occurred when trying to open native Office docs. This is no longer a problem. Compatibility is almost near perfect; your colleagues will be able to tell that any of your docs where authored outside of Microserfdom.

• SeaMonkey

  MOAB (Mother of all Browsers). I know that Firefox and Thunderbird are getting all the press at the moment, but harken back to the heyday of Mozilla and you will appreciate the full-featured power of this next-generation online suite. It has all the advantages of its derivative offspring without the misfeaturism which mars the beautiful simplicity found in SeaMonkey. If you value stability over showy spectacle, this is the browser for you.

Now, onto my litany of grievances. As I stated, OS X is not perfect. The following represent a smattering of the issues that I've encountered :

• Bluetooth

  The idea of a personal area network has been realized with Bluetooth enabled devices and although I love the idea, having used the Wireless Keyboard and Mighty Mouse, the Apple implementation leaves something to be desired. Firstly, their is an ongoing issue regarding bluetooth and sleep/suspend modes. Despite numerous system updates, this issue remains unresolved. Suspending OS X and having it detect the presence, or lack thereof of bluetooth devices should be seamless. Given the relatively small variation present in Apple's product offerings, I would think it would fairly easy to standardize hardware used in their machines. You would think that having a relatively homogeneous hardware platform would translate in to better driver support, but not so. Bluetooth is one of the few components on OS X that can get the system to kernel panic requiring a hard restart.

  In addition, the latest wireless 802.11n updates have introduced signal jitter on connected bluetooth devices when transferring at high throughtput. Since Bluetooth and Wireless BGN share the same frequency spectrum (2.4 Ghz), I suspect the two are related. You ask, why not use the 5Ghz wireless frequency? This brings us to the next of my gripes ...

• Airport Extreme 802.11.n support

  Theoretically, 802.11n should allow for roughly 500Mbs of wireless throughput in either the 2.4 or 5 Ghz frequency spectrums. This is 5x faster than 100Mb wired Ethernet, or approximately 10x faster that wireless 54G. So much for theory. I bought the Extreme in the hopes that it would improve the speed of my wireless connections by being compatible with the wireless 802.11n adapter in my MacBook Pro. It has improved in speed, if at a more modest 2x increase. Not only that, the latest updates seem to have throttled the top speed to 300Mbs. The network manager reports this throughput despite the fact that my machine is no further than a few inches away from the base station. There are other annoyances as well: Why is channel selection disabled for 802.11n? Why does 802.11n fail at relatively short distances when using the 5Ghz spectrum in my 800 ft/sq space? It literally allows for connectivity only as far as the room where my access point is located Why are there no diagnostic tools to help determine possible causes of signal interference?

  I understand the complexities of creating OS device drivers, but I wish Apple would perform more extensive QA testing before rolling out "fixes" which actually degrade either performance or functionality. At the very least, they should be forthcoming in the compromises between various updates.

• Lack of network signal strength / wireless info

  This is a no-brainer, and for the record I am already aware of third party utilities that accomplish this task. The problem I have is that this feature should be something built into OS X. A bit of browsing on the Apple Forums reveals a pattern of wireless connectivity issues related to the installation of third party utilities. It stands to reason that avoiding anything that touches the network stack is a "good thing" given the prevalence of these types of issues. At the bare minimum, Apple should introduce a wireless connectivity tool that allows you to view networks within range and their associated security info and signal strength.

• Hidden files in Finder

  The ability to control which files are displayed in Finder should be customizable. It should not require a Finder/Terminal hack to create/use Unix . files. I understand the necessity to isolate the less experienced user from the Unix internals, but this should be customizable through some mechanism such as a defaults setting. There are a number of legitimate uses for . files and more experienced users should not have to suffer for "perceived" security considerations with no real merit.

  This "feature" was particularly problematic for me when I attempted to relocate my iTunes library to a Samba network share. Unknown to me, OS X stores directory metadata in a readwriteable file .DS_Store. If you prevent the creation of . files on the Samba share, OS X will fail mysteriously. Just keep in mind that . files are useful for normal system operation as well.

• Runaway processess

  There is no way other than having top run in a Terminal or the Activity Monitor running constantly in the background to know that a system process has failed unexpectedly and is running your CPUs at full throttle. The more astute will notice the Chernobyl-Like-Heat radiating beneath sweaty palms, but the common user will only notice the disastrous after effects of such a system.

  The number of heat related posts in the Apple Forum are staggering. Being able to identify and terminate a runaway process is a sure way to maintain the longevity of your system. There's nothing like having full CPU utilization and both fans running to drain the battery prematurely as well.

  If the two system monitoring alternatives presented above don't appeal to you, I recommend downloading the iStat widget. Not only can you configure it display the most resource intensive processes on you machine, but it also provides a host of additional information such as the temperature of various thermal sensors in your machine.

• Half-baked updates

  This is a peeve which I carried over from Windows and did not expect to have to deal with in OS X. If updates are to be release for public consumption, at the very LEAST, some regression testing should be done on the updates so that at the minimum they do no harm to an existing system by destabilizing it, and at most, they actually fix the problems they purport to fix. Lateral progress is to be expected sometimes, but retrograde progress is a definite no-no.

• Safari

  What can I say? As the default browser of OS X, I found Safari somewhat underwhelming. There is nothing here that you can't get with the more capable SeaMonkey. The truly, truly crappy part about Safari is that the system wide browser settings (such as the default browser) are embedded in the Safari Preferences. How about putting these settings in the System Preferences so that we can remove the necessity to have Safari run on the system at all?

• Mail

  This application is far from being useful and is targeted towards the light-weight/casual email user. Its perennial speed issues as well as its inability to deal with large mailbox sizes (think in the tens of thousands of messages) makes this probably the most useless OS X app for me. I didn't even last a day with Mail before deciding that SeaMonkey was the way to go. Unless there is a compelling reason for Apple to develop Mail and Safari, users would be well advised to avoid them entirely.

Apple is not Microsoft and for that, I am grateful. But if they continue in this vane, the quality that distinguishes them may become diluted and eventually lost. As Apple users, we must insist on a level of hardware/software integration that exceeds user expectation by simply working. I realize that this is idealistic, but I believe this to be the core of the public's perception of Apple, and should not be disregarded lightly. If Apple is to have any chance to grow their market share in the personal computing segment, they should avoid disabusing their relatively tolerant user base who have paid a premium for their product on the assumption that they would be entitled to a hassle-free computing experience.

The introduction of Leopard is a watershed moment in computing history which will mark a turning point in how we view operating systems and their role in the computer-application-user triumvirate. Instead of the OS being simply a platform to support the computer-application relationship, focus will shift to its role between the user and the application.

In this context the operating system will change from being a hardware enabler to being an application enabler. In this regard, OS X with its Unix heritage is perfectly suited to this role.

The beauty of OS X is that it does not pander to the lowest common denominator with regards to the User. Inexperienced users value discoverability. Advanced users value transparency. The inexperienced want to be able to intuit the operation of an application without having to read lengthy manuals. Advanced users want an application that does not prevent them from using efficiencies gained through experience to accomplish tasks.

There is the implicit acknowledgment in OS X's design that the technology quotient of its users will run the gamut. The operating system's role is to provide a suitable framework to orient the user experience regardless of his/her technical ability. By not forcing a choice, OS X meets the primary requirement of "Users"; usability unencumbered by singular vision of compromise. By challenging the viability of a single OS monoculture and embracing the strengths of both Linux and Windows, Leopard reveals the truth about the power of multiplicity and diversity.

The principle of focusing not only on "what-it-can-do" but "what-you-can-do-with-it" is something which pervades Unix and indirectly OS X. Utility and usability. Given the right combination of hardware and software, anything is possible but it is the intent that characterizes the system. I'm inclined to believe that the harmony between utility and usability is the source of intangible "quality" that pervades OS X. Unix in one incarnation or another has evolved since in inception 30+ years ago as a platform for innovation. This quality is why Steveo's had his heart set on having a commercially viable Unix alternative. With Leopard, I believe that his wish which began with NeXT has finally been realized with Leopard.

How do OS X's rivals fare in comparison?

Although OS X shares a lot of functionality found in most commercial Linux distributions, Linux by and large is still focused on the computer-application relationship. With the exception of the Digerati, this focus prevents true widespread adoption. Until the usability gamut is addressed without diminishing its strengths, it will be relegated to infrastructure products and not the desktop. Although these weaknesses are being addressed by various commercial Linux vendors (SUSE / Redhat / Ubuntu), Linux as an application platform is still far from maturity. Until Linux moves away from mimicking equivalent functionality found in its competitors and rids itself of its preoccupation with superfluous eye candy, no true innovation can occur.

Surprisingly, Linux's weakness is also Microsoft's. By constantly pandering to self-serving industry initiatives (i.e. DRM, Genuine Validation, etc), they too have focused disproportionately on the utilitarian aspects of the Vista at the expense of the User. Vista is simply Microsoft's expression of innovation envy without the innovation. Is their any justifiable reason for upgrading from XP to Vista other than to enforce the digital rights of media companies and mitigate Microsoft's less than stellar security track record? Not from the initial reports of early Vista adopters. It is simply hubris that a company can manipulate is user base into paying for the privilege of beta testing its software. Despite the large (contractually enforced) install base, Vista is shaping up to be another Windows ME debacle for Microsoft.

• Installing Postgres 8.2 on Leopard (includes system account!)
• Starting PostgreSQL 8.3 through launchd on Mac OS X 10.5.1 Leopard
• Installing PostgreSQL on Mac OS X
• PostgreSQL on Mac OS X

I spent a bit of time trying to figure this out, and instead of having to relearn this painful lesson in the future, I thought I would preserve my tribulations for prosperity. There were a number of resources that provide partial guidance in discovering the solution (which I've included here), but here is the solution laid out in full:

---

1. Install Fink

If you haven't already, install Fink. As per the Fink install instructions, you will also need the Apple Developer Tools. The package tools which come with Fink are second to none. To pull down the latest version of a package, simply install it using FinkCommander from source or binaries.

2. Install PostgreSQL

At the time of writing this, the latest version of PostgreSQL is 8.2.6. It is highly recommended that you use the latest available stable release of PostgreSQL as it usually includes a number of security and stability fixes which are a prerequisite for maintaining a stable, secure installation. To install PostgreSQL from the command line, you could also run the following command:

$ /sw/bin/fink install postgres

Simply follow the prompts to complete the installation.

3. Create PostgreSQL account

I found that previous versions of PostgreSQL left some stranded permissions on my system. Unfortunately, none of these mechanisms allowed me to sudo into the postgres account. Given the change in user management tools in Leopard, it is recommended that you use the new Directory Service commands provided in 10.5.x to add the PostgreSQL user to the system. If you have a previous version of PostgreSQL installed that you are not currently using, I recommend removing it and any associated users from the system. Exercise caution doing this if you have an existing installation as it has the potential to loose your data if you are not careful. As the PostgreSQL account used to manage the database should behave like other system accounts, you will need to associate with an unused user and group ID which is less than 500. This will prevent it from being displayed in the Login Screen. I modified Pat Maddox's instructions to determine which whether or not a particular ID is currently in use by issuing the following commands:

$ dscl . -list /Groups PrimaryGroupID | awk '{print $2}' | grep user_id
$ dscl . -list /Users UniqueID | awk '{print $2}' | grep group_id

where user_id and group_id are the IDs you would like to check. If the command returns a value, then the ID is already in use. In most cases, user/group ID 499 is available for use. The rest of this example uses this for the PostgreSQL user/group ID. To create the PostgreSQL group account issue the following command:

$ sudo dseditgroup -o create -i 499 -r "Postgres Admin User" pgsql

To create the PostgreSQL user account issue the following commands:

$ sudo dscl .
> create /Users/pgsql
> create /Users/pgsql UniqueID 102
> create /Users/pgsql UserShell /bin/bash
> create /Users/pgsql RealName "Postgres Administrator"
> create /Users/pgsql NFSHomeDirectory /usr/local/pgsql
> create /Users/pgsql PrimaryGroupID 102
> create /Users/pgsql Password *
> quit

4. Configure directories

This is where Maddox's instructions (based on MacPorts) and mine diverge. Fink typically installs packages under the /sw directory. The main PostgreSQL binaries should be located in the bin subdirectory with the database files installed in the var subdirectory. The database directory usually has a version suffix. Once you've located the database directory, you will need to reassign it's ownership to the newly created pgsql user. I usually provide a symbolic link which points to the most recent version installed on the system to provide a short-cut as well as indicate which version is being used (when more than one version is installed).

$ sudo chown -R pgsql:pgsql /sw/var/postgresql-8.2
$ cd /sw/var
$ sudo ln -s postgresql-8.2 postgresql

You will notice that PostgreSQL was not installed in /usr/local by Fink. This avoids potential conflicts which might arise with other 3rd party packages installed on the system. You will also notice that when we created the PostgreSQL account, the NFSHomeDirectory attribute was set to /usr/local/pgsql. This was intentional. By creating a symbolic link from this directory to the one created by Fink, we can integrate the Fink installation seemlessly in with the rest of the system without introducing conflicts or dependencies. I usually prefer relative paths in this circumstance:

$ cd /usr/local
$ sudo ln -s ../../sw/var/postgresql ./pgsql

By pointing to the original symbolic link, simply changing that link will automatically update the NFSHomeDirectory attribute associated with the PostgreSQL user.

5. Access caveats

Resist the urge to associate a password with the pgsql user. Using sudo to assume privileges will allow you to limit who can access your PostgreSQL installation locally. If you require remote administrative privileges, configure accounts through PostgreSQL directly instead of delegating privileges to system accounts.

6. Configure database

At this point, you should be able to switch to the pgsql user. Check your identity and privileges:

$ sudo su - pgsql
$ whoami
$ id pgsql

To initialize the database issue the following commands:

$ /sw/bin/initdb -D /usr/local/pgsql/data -E utf8

This will create the initial postgres database which will support UTF8 (for international character sets). Once the database is initialized, you can start it by issuing the following command:

$ /sw/bin/pg_ctl -D /usr/local/pgsql/data -l /usr/local/pgsql/logfile start

To stop the server:

$ /sw/bin/pg_ctl -D /usr/local/pgsql/data stop

To enable the procedural language for all databases derived from template1:

$ /sw/bin/createlang-8.2 plpgsql template1

You should now be able to connect to the postgres database:

$ psql -d postgres

7. Security

At this point you might want to spend some time securing your PostgreSQL installation. There are a few pointers in the Installing PostgreSQL on Mac OS X article, but as with all things Postgres, please refer to the Manual for definitive answers regarding security. One of the few things I did to secure my installation was to modify the data/pg_hba.conf file to support a default administrative user that required password authentication. This prevents a local user from automatically gaining access to the database by compromising the PostgreSQL user account. This required first assigning a password to the PostgreSQL user account. This is accomplished by connecting to the newly created postgres database and executing a SQL statement which assigns the user a password:

$ psql -d postgres
#postgres=# ALTER USER pgsql PASSWORD 'secret';
#postgres=# \q

where secret is the password that you've selected. The password itself will be stored as an MD5 checksum so that when you authenticate against the PostgreSQL server, it is never transmitted in plain text. Next, the data/pg_hba.conf file needs to be modified to support authentication for the PostgreSQL user account. Use an editor to add the following lines:

local    all         pgsql                             md5
# IPv4 local connections:
host    all         pgsql     127.0.0.1/32             md5
# IPv6 local connections:
host    all         pgsql        ::1/128               md5

After these changes, a server restart will be required:

$ pgctl -D /sw/var/postgres/data reload

When logging into the default database you should now be prompted for a password. Be careful if lines such as the above already exist in your configuration. You may need to replace them in order for this to work.

8. Tuning

As this database is going to be used for development purposes, a lot of cruft will accumulate in the database as tables are create and destroyed and data is modified. To save the trouble of having to automatically clean up the database, you can modify the data/postgresql.conf file to support auto-vacuuming to recover space in the background. The following lines should be placed in the file:

stats_start_collector = on
stats_row_level = on
autovacuum = on 

As the postgresql.conf file is installed with defaults on a fresh PostgreSQL install, you may need to simply comment out these lines.

9. System service configuration

The last step requires configuring the PostgreSQL install to start and stop as other OS X system services do. This will require creating a StartupItem entry for PostgreSQL. These basis of these instructions are courtesy of Robert Crews and wmaus. First create a directory to store the initialization script for PostgreSQL:

$ sudo mkdir /Library/StartupItems/PostgreSQL

Create the PostgreSQL script under this directory with the following code:

#!/bin/sh

# Source common setup, including /etc/hostconfig
. /etc/rc.common

FINK_BIN='/sw/bin';
PGSQL_HOME='/sw/var/postgresql';
PGSQL_USER='pgsql';

StartService ( ) {
    # Don't start unless PostgreSQL is enabled in /etc/hostconfig
    if [ "${POSTGRESQL:-NO-}" = "-YES-" ]; then
        ConsoleMessage "Starting PostgreSQL"
        sudo -u ${PGSQL_USER} ${FINK_BIN}/pg_ctl \
            -D ${PGSQL_HOME}/data \
            -l ${PGSQL_HOME}/logfile start
    fi
}

StopService ( ) {
    ConsoleMessage "Stopping PostgreSQL"
    sudo -u ${PGSQL_USER} ${FINK_BIN}/pg_ctl -D ${PGSQL_HOME}/data stop
}

RestartService ( ) {
    if [ "${POSTGRESQL:-NO-}" = "-YES-" ]; then
        ConsoleMessage "Restarting PostgreSQL"
        StopService
        StartService
    else
        StopService
    fi
}

RunService "$1"

Obviously, replace the constants at the beginning of the script to match your installation if they are different from the ones here. There is one step that is not described in the Crew's instructions which is required to allow the SystemStarter to manage the service. A PostgreSQL.loc file in the same directory as the script which contains the path to the postgres binary and provide the execution context for pg_ctl must be created. In the case of a Fink installed PostgreSQL installation, this file should contain the following line:

/sw/bin

Without this .loc file, the SystemStarter will throw an error when invoked.

The last step requires adding a PostgreSQL entry in the /etc/hostconfig file. It's always a good idea to save a copy of the original config file before modifying it. Add the following entry to the file:

POSTGRESQL=-YES-

This will ensure that PostgreSQL is started when you boot your machine. You can also manage the service using the following commands:

$ sudo SystemStarter stop PostgreSQL
$ sudo SystemStarter start PostgreSQL
$ sudo SystemStarter restart PostgreSQL

---

Drop me a line if you find this useful or if you find any mistakes in the method outlined above.

If you are using Ruby in OS X, I suggest NOT managing it through Fink. OS X comes with Ruby pre-installed. All that is required is to update and install a number of gems as per the Apple Developer Instructions. The instructions also cover configuring Xcode to use for Rails development. After trying both Aptana and the Rails plugin for Eclipse, I suggest the path of least resistance. Xcode is particularly elegant and uncumbersome.

# sudo gem update --system
# sudo gem install rails
# sudo gem update rake
# sudo gem update sqlite3-ruby

Building the PostgreSQL gem requires an additional step as the installed database is managed through Fink (thanks to Roy Hooper fom this one):

# export PATH=/sw/lib/postgresql82/bin/:$PATH
# sudo env ARCHFLAGS="-arch i386" gem install --remote postgres

Modify the environment.rb file to reflect the version of rails installed above. For example, the default scaffold built by the rails command assumes that the version is 2.0.2. To prevent the following error:

Missing the Rails 2.0.2 gem. Please `gem install -v=2.0.2 rails`, 
update your RAILS_GEM_VERSION setting in config/environment.rb for 
the Rails version you do have installed, or comment out RAILS_GEM_VERSION 
to use the latest version installed.

I had to append .9097 to bring the rails version to 2.0.2.9097 for this rails release at the time of writing this article. This will most likely be different for newer versions of rails. The version option for the rails command returns version 2.0.2, but the issuing the following command will return the release number :

# gems list

If you update your gems, you will have to update the environment.rb file with the release version.

I also do not recommend using Aptana. Although it is tempting to use an integrated IDE, there is a bunch of functionality that is buggy (such as the Generators).