R – How are objects defined in an XIB file and how are they created at runtime

First, placeholder is a better word than proxy here.

Normally when you have an object in a NIB/XIB file it means that loading the NIB file will create that instance. The placeholder objects are objects that will already exist when the NIB file is loaded, and they appear inside of the NIB so that you can make connections between the objects that will be created by loading the NIB and the objects that already exist.

The file's owner, first responder and application are all placeholders.

The file's owner is placeholder for the object that will load the nib. All of the NIB loading methods take an 'owner' parameter. When you make a connection with the File's owner, when it's established at runtime, it will be connected to the owner object passed in to the nib loading method. Many UIKit and AppKit classes invoke the nib loading methods for you. NSApplication, NSViewController, NSWindowController, UIApplication, and UIViewController all load NIB files on your behalf. When they do that they pass self as the owner parameter to the nib loading methods. That's why when you use a view controller or a window controller you set the file's owner to your subclass and make most of the connections between your views and the file's owner.

The NSApplication instance is a simple placeholder for [NSApplication sharedApplication]. That's a global singleton and the icon in Interface Builder represents that global singleton. Loading the NIB file does not create a second NSApplication instance. By contrast, when a NIB file contains a window, if you load it a dozen times, you'll have a dozen window instances, but still one NSApplication instance.

The first responder is unique. Connecting an action to the first responder means that when the action is fired, it should dynamically be sent to the responder chain. The responder chain typically starts with the focused view, and continues up through the view hierarchy and includes some controllers and delegates. Each object in the chain gets a shot at handling the action. Menu items work great with the responder chain. If you had a menu item for "Make Bold" that is supposed to make the currently selected text bold, you might start by hooking that up to an NSApplication subclass, but then you'd have to know all of the situations that "Make Bold" applies, and how to handle them. A text view and an editable web view would probably need different code to handle "make bold" and bottling this all up in one object would get quite complex and wouldn't be very extensible. Instead you could connect the "Make Bold" menu item's action up to a makeBold: action on the First Responder. This would mean that when the menu item was selected, the focused object, or one of its parents that responded to makeBold:, would get the makeBold: message. Now many classes can implement a makeBold: method and respond to this menu item when they're in focus.

The rule of thumb I use, is auxiliary objects can be put in if they are related to the UI being defined in the nib in some way - so either proxy objects that hold references to elements or actions that get triggered.

Proxy methods all get wired in when loading the nib and pass in the already created object via the userinfo dictionary you can optionally pass in along with the nib name.

There's nothing at all wrong though I think, in a purely logical nib that would use the target/action system to wire up multiple proxy objects. That might be easier than doing all the wiring in code, though I've not seen that used in practice.