Uh, no, that's the hint that you need. It's irrelevant what happens in the
CDN.
It's actually far from irrelevant. The CDN is the major reason "streaming" services are viable in the first place. Without the CDNs we'd be back in the real streaming era of RealPlayer et al, right before that small company Akamai saw a business opportunity...
You're still going to stream it to your player.
Nope. Your player reads it from a growing file on disk. If you only want to concentrate on that part and call it streaming, then you've already lost the whole transporting across the network part, including controlling what gets streamed while it happens.
If it starts playing
before you finish downloading, you're streaming.
Still incorrect. That's just playing a partially completed file on disk.
It's "streaming" in the most primitive way possible. For example, you can't seek forward and play another part of the media until you've waited for the full file up to that point to be downloaded. True media streaming technologies allow the player to seek forward, *and not get the bits in between*. Basically, the server doesn't send any unnecessary bits, modulo the encoding method.
In your "disk as the media server" analogy, you would want to have some way to access the file randomly - that's only possible when the intermediate data has already been stored.
If you have to wait for the whole file to download before
you can watch, then it's not streaming.
There you go. We seem to be in agreement after all. If you have to wait
for 90% of the whole file to be downloaded before you can seek to the 90% position, then it's not streaming.
There probably are still video
services like that, but I don't know of any.
There's an easy test you can perform to find out: stream a large file (large enough to take some time to download) and right at the start, have your player seek to somewhere near the end. If it starts playing the end straight away, it's probably a streming player, but if you have to wait a while, then it's probably a downloading player.
Whether the video gets downloaded into a buffer in memory or a buffer in a
disk file is completely irrelevant to the question of whether you are
streaming. It only speaks to the issue of how it is done.
Not so. I argue that the difference in technology is sufficiently important to the end user experience that without downloading tech, streaming media business models wouldn't have taken off - as they didn't in the early days when this was tried.
Ultimately, my point is that "streaming" as some of you use the word is a marketing term, useful to give the impression that data arrives on demand into the player and disappears as soon as it has been used, whereas this isn't the reality. True streaming like that is certainly possible and there are servers that do it,
but mostly it's downloading files and hiding them. And I happen to like finding them and maybe processing them in ways that work for me.