For me this is all about risk and threat level analysis. At the core of the problem is trust:
* do I trust company privacy policy? (me: no)
* do I trust the company not to misuse my data? (me: no)
* do I trust the front-end app not to siphon out my data to the company? (me: sort of - I am able to check, and many others do too, so there's a level of public scrutiny)
* am I fundamentalist not to use their app, even if it's OK as far as anybody can tell? (me: no, other apps are either worse, or just don't work)
* when it comes to workflows where I legally *must* compromise my identity/privacy (e.g., payments), it does not matter what I do, law & regulations overrule. Implication: I will use Google Play/Apple Store, because there's no way around if I want to pay developers/companies for good products.
* do I care about long-term data availability.portability (hell yes!) -- here is your silo avoidance.
Summarizing this way, I concluded that:
1. I do want to run my own backends, while
2. I want them to use open and long-term standards (Maildirs, SMTP, etc., WebDAV, CalDAV, CardDAV for sync)
2. using convenient front-ends and relying on 3rd parties (public, foss & white-hat community) to keep me semi-safe from siphoning my data elsewhere; and
3. glue the whole thing with independent/ideally open source middle-man plugins/conduits.
My design (here I agree with @aku):
1. trusted e-mail provider (almost self-hosted),
2. ownCloud/nextCloud/private NAS for file storage + calendar and contacts backends,
3. independent e-mail apps (mutt, ProfiMail on Android)
4. Google Calendar on Android + Mozilla Seamonkey/Lightning on Desktop or Google Contact app for front-end (no backup to cloud, no sync to Google!) -- but here you could go more convenient with KOrganizer with
5. CardDavSync/CalDavSync apps in the middle to avoid Google's prying hands.
Works for me since many years. And most importantly, it works for everybody around me too, as if they chose to use Google/Apple stuff, they still can because of open standards. This is not a problem of e-mail, but primarily of calendar.
