On Mon, Jul 06, 2020 at 11:15:01PM +0000, таракан wrote:
In the strict P2P concept, your friends or contacts may well get annoyed if you keep hitting their phone with requests for the same image for example.
Caching and decentralized content are 2 different things I guess
Well this is a mailing list of people who, more or less, are using technology and/or cryptographic science to ensure privacy for "anyone"...
In such a task, caching appears futile, unnecessary and an obvious risk. Therefore, better not to even considering it, that solves the issues.
I develop an embedded system for a secure communication station. I want everything to stays transient, to be erased as soon and as fast as possible. I don't want that it is possible to scan the memory to intercept any variable, deciphered stuff etc...
In that communication station, communications can be eventually extremely scarce, internet very slow in extreme environments with all sort of modems involved.
Since this is a secure communication station, the last thing I want is to cache anything, simply because that is a secure system. The received data are stored in encrypted external components.
Etc...
Sounds like a relatively high security setup - this would likely be beyond the average user. This distributed/decentral content is interesting - I've been thinking of "cache" as the local node's "contribution" to the distributed P2P content store. In this context, some devices might be considered bridge or entry nodes as Tor call's them, but once a user start's viewing content, I had not imagined that their device for reading and responding would _not_ cache files, at a minimum for a session. In this early "explore the space" trial, I'm looking at the workings of such an app, so I can hopefully get a better understanding of the overall network and user needs. Also, despite the monumental first mover advantage that the incumbent _centralised_ social media giants have over everything, I still hope that one day something fully decentral might by some fluke of providence be at least marginally successful enough to provide an "off the grid" alternative. So in this decentral context, all "servers" are legacy or at least they "cannot be relied upon". So how do we design for this is the next question - "servers" must be as trivial as a hash tag, and this has implications. Each end user desired node is a use case - there is evident value e.g. in separation of one's Internet/overlay entry node, and one's browsing node. But I think that the use case of a cache-free browsing node needs more consideration: How do we avoid a node "hitting up their peer nodes for the same media object repeatedly"? In P2P, nodes must collectively provide the content store - how do they do this? Perhaps the "local distributed partial store", if a node has one, is never accessed by the local node? Possible I guess - had not thought of this, and there's no reason this could not work if it's useful, but it might bring pressure to simplify "offline replication into a secure vault/ encrypted store, for offline browsing of your past conversations". And once you have an offline vault sink, say a mounted Veracrypt volume, then why not also use this to store avatar icons? Once again, we're back to a form of local cache, albeit a little more complex for the user to set up - in that case, the pocial media client could be split into separate apps if it makes user's feel safer (may be there's already a Veracrypt for Android?), or if you have a local "browsing" computer, perhaps a "favoured peer node" (on your workstation) could serve all cache requests when that node is contactable so that your "browsing" node never caches? Once a use case is properly understood, it can be properly provided for... (BTW, in my initial concept tests, I'm implementing a social media client test ui, in order to understand the issues. Networking could run over Tor, I2P, clear net, or some future overlay.)