Is Dojo's load-as-you-need it structure actually a performance improvement? For me, at least?
My company's website is going to switch to IBM Websphere, which primarily uses Dojo. My company's very concerned with page performance, mostly in terms of "seconds to page load". As a result of that, the directive we've been given is "minimize hits to the server", so with our current website we aggregate all our .js files before the promotion to production.
But that directive is basically becoming Law, now, so if I were to argue against it, I would need a very good rationale for it. I've been unable to find anything in favor of the load-as-you-go method except "it's a good idea" and "loads only when you need it" (the latter is really only based on the former, as far as I can tell).
And then, if I were to flatten everything out into a single file, I wouldn't be able to use dojo.require() statements at all, would I? (the idea being, if I could have the development side split by module to make the organization more rational, but then have the production version a single file, but then dojo.require() would no longer make sense there, and then I have an increasingly complex situation where I would need the build to do some invasive things to the javascript to package for production.)
Please resist the "it depends" answer. The best practices docs I've seen (Yahoo, Google, etc) pretty much just say "reduce page loads" and don't have much "it depends" about it. But Dojo's framework seems so definitive about its approach, I'm wondering if there's a more persuasive argument for it.
