tl;dr How should an application handle windows for content-savable-to-disk (sort of documents) behave under an OS that has menu-bar-per-window (Windows, Gnome Shell) and under an OS that has single OS-wide menu bar (OS X, Unity)?
My company has an application that creates two different kinds of documents: an 'activity' document—configuring report parameters—which then spawns one or more 'report' document—graphs and other read-only data that the user may save to disk. The application currently only runs on Windows, and currently uses a browser-like interface: the application has a single window (single menu bar) and each new document (of either type) creates a new tab.
I'm specifying the port to Ubuntu and OS X now, and trying to decide whether or not to create one window per activity and one window per report, or stick with browser-style interface.
The Impact of Each Decision
On an OS with a single menu bar (OS X, Ubuntu w/Unity) it's fine to create one window per activity and report. The user may choose to close the last window and still the app remains running, with menu commands to open an old document or create a new activity.
On an OS with one menu bar per window, closing the last window must quit the application. Thus, if I have a window per document, I have to create some window (a new activity document) when the application is launched.
If I go the browser-style route, I am flaunting the Apple Human Interface Guidelines regarding document windows (and losing out on nice features like the proxy icon in the title bar). Further (unless I go a horrible MDI-style path, or the browser-like ability to move tabs to new windows) I lose the ability to visually compare reports side-by-side.
Are there any guidelines for this on Ubuntu, especially when programming an app to be used between the two different application menu paradigms? What heuristic applies when deciding between browser-style tabbed 'documents' versus window-per-document?