You can also just upload, and then not share the image (it's a separate flow after the upload that you can skip.) The default setting for an uploaded photo is "not shared".
• If true, it's a really fucked up UI/UX. Because the two actions have absolutely no need to be associated.
• If not true, it's a really fucked up UI/UX. Because in two years of using photos, I haven't sorted it out, and I routinely see people sharing individual photos to their streams (I occasionally ask "context") to find that they're assembling an album of some sort.
Again: the photo-sharing is pretty much useless on account of that, and Imgur shines by comparison.
If you want persistent storage you control, buy an S3 share or host your own. Frankly, broadband access is to the point the latter is viable for personal accounts. I'd like to see some sort of P2P distributed cache which shares load as well.
I'm not sure what you mean. If you don't share a photo when you upload it, you can share it any time later. (In particular, that's how auto-upload works.) If you share it, and then you delete the post where it's shared, it goes back to not-shared. That seems pretty straightforward.
It doesn't seem like a surprising UI to follow an upload with a share when you're sending photos to a social network. I'm not even sure Facebook gives you that choice. The two actions don't have to be associated, but they certainly need to be associated when you consider that folks usually upload photos to share them.
I don't think G+ Photos has been advertised as a "persistent store you control". It's a social network that has an auto-upload feature to make it easier to share images and backup shots taken on your phone. "Hosting your own" is not going to be a substitute for 95% of real-world users.
Honestly, the G+ photos UI is sufficiently fucked up that I've never been able to sort any of this out.
I see Google's offerings as suits my needs. And Photos hasn't offered that.
I disagree on HYO, because reasons. Persistent broadband, a $25 device, and 5 watts will get you a server. The software's free. Some form of distributed federated caching gets you redundancy and load balancing. Search is the tough nut, though there are a few projects which have been working at that (e.g., YaCy) for a while now.
Otherwise it's just protocols, autoconfiguration, and adoption.