Hacker News Reader

Comment by PhilWright | Hacker News Reader

The problem with your personal data store is that it will be impossible to stop the large companies like Google/Facebook from simply copying and caching a copy of all the data they need to access. You give them permission to handle your email and calendar information and bingo, they have a copy of your entire set of emails and appointments. If you use multiple services from Google/Facebook then they will end up with a good copy of your entire dataset and can mine it of value just like they already do.

View on HN

CouchApps had a better vision for the web. You would run a local CouchDB and replicate the app next to your data, rather then giving the app access to your data while hosted somewhere else. It has the added benefit of working offline.

Some sort of cross-domain request blocking could then prevent the app from stealing your data.

Thanks for mentioning CouchApps (based on CouchDB, or the scalable BigCouch). I haven’t thought about CouchDB as a platform for a long time, but it does hit a sweet spot for local data, distributed replication. Seems like a good basis for a Decentralized Web on a small scale where you share with a small number of people.

That's correct, and if the companies can't see the data, then the data is really just an extension of client side storage, and therefore it can't be trusted, because you can never trust the client.

I think that's the aim, not the failure of this plan.

The key difference is that instead of companies storing, owning and exploiting your data, the user maintains their own data store and companies can then access and exploit it.

It's decentralised in the way where the data lives and is controlled by, not in how or where the data is used.

There is the problem where companies could continue to use your data when you no longer let them access to your data store - but that in effect is at worse unethical and at best not legal these days.

The datproject [1] had an interesting approach to this. Distributed data stores that you control and are addressable via hashes, with client-side apps that work 100% in browser just by pulling in the distributed copies of that data (via its hash) - no backend slurping it up.

Of course, it is not impossible for the client-side app to send the data back up to the server. Even if encrypted somehow once decrypted in the client for display to humans then its hard to protect that plain-text data. There were also some other open questions like if my distributed data is distributed, how can I "take it back?" I dont think there is a mechanism for that yet - some sort of TTL might work there though, if you can somehow bake it into the data/hash itself to avoid bad clients from ignoring TTL values before deleting/archiving.

1 - https://datproject.org/

On the other hand, what would stop the rise of email UX companies that provide a PAID service that links to your email with the explicit promise of never caching it? They wouldn't need to mine it to make money as google does, as they are being paid, even though the per user cost payment could be quite low to support a profitable email UX company.

Google does not mine/scan gmail emails for advertising purposes.

https://www.nytimes.com/2017/06/23/technology/gmail-ads.html https://www.engadget.com/2017/06/23/google-wont-scan-your-gm... https://www.wired.co.uk/article/google-reading-personal-emai...

Many more sourcews just a short duckduckgo search away.

How could you prove/trust that they never cache the data? I think the only way is if the "app" actually ran on your PDS (Personal Data Store) instead of somewhere else. Of course that brings other challenges.

Ideally you'd pay them and that would be part of the contract. Otherwise demand a legally enforceable promise (via promissory estoppel).

I also think there is a need to figure out how to enable community-controlled SaaS platforms in addition to this. LibreOffice, for example, has essentially released an online office suite. But it has decided not to actually operate and offer this suite to the public in ready-to-use fashion. From what I understand, it is just too difficult and resource-intensive to do that. So it's up to companies running paid platforms to do it (or you can spin it up on your own server/instance and run it yourself). There's nothing inherently wrong with this. But it seems to me that there's a "next step" to take by figuring out how to enable fully community-controlled platforms so that a project like an open source G Suite (running at scale with an iron-clad privacy guarantee that is backed by community audits) that you can just go create an account on could become a reality one day. This seems like it would require a non-profit organization akin to Mozilla. How great would it be to have a community-controlled non-profit organization operating a trusted cloud platform, perhaps even audited by a group like the EFF? Very challenging, for sure. But it seems humanly possible.

This means that apps/services will need to come with a self-hosted option. This is technically doable (at least for email/calendar), but it implies that vendors will need to re-adopt product-centric business models, solve distribution issues, etc.

Reply on news.ycombinator.com