To try to summarize this on a high level:

A simple new key type would allow us to have spam-proof forums using a single 
key queue, like on Frost, as opposed to having to poll hundreds of identities 
as in FMS and WoT. The principle is to allow posting as long as your key has 
been signed by the moderator. If you start spamming he can change keys and 
re-sign everyone else's identity. Such groups would be of limited size, but 
likely usable for forums in practice, and there are various ways to make this 
scale better.

A more general-purpose solution would be a key type that allows for arbitrary 
"posting rules", written in a non-turing-complete language. This would improve 
efficiency for such forums, and would allow a wide range of other 
functionality, including hashcash-protected keys and revocable keys, and 
probably lots more. IMHO we should go for the more general purpose key type, 
but as Sadao has pointed out, for forums there are workable solutions with the 
simpler type. (Elliptic curve crypto would likely be very helpful as it has 
much shorter keys)

We can also use these tools to massively speed up Web of Trust-based chat 
systems, so the fact that you need to trust a moderator (or group) isn't 
necessarily an issue.

IMHO such a speedup would have massive advantages for usability:
- Latency would be reduced dramatically. Even real-time chat like FLIP could be 
spam-proof.
- Messages would download much more quickly on joining a board / installing the 
chat client.
- The load on the network created by message systems would be greatly reduced.

IMHO Web of Trust based chat can scale adequately provided that:
- We use the above optimizations
- We poll mainly the identities that regularly post on our subscribed boards
- We have some sort of per-board announcement system so that we can see 
newbies, even if the regular posters are ignoring them. This might need 
hashcash, I'm not sure yet...

Sone can of course scale well in the same way as Facebook, even without these 
changes - by only fetching updates from your friends, and providing an out of 
band means to contact you. (The same mechanism as per-board announcement above).

Distributed searching could use the new key type, but there are scaling issues.

Full-blown distributed data structures are possible if we add a further key 
type that allows the posting rule to fetch a limited number of other keys. 
Which could greatly accelerate e.g. searching by allowing multiple posters to 
maintain one index, which could then be efficiently searched or downloaded. But 
that's something for the future.

Conclusion: The list of "impossible" things we need to do to make Freenet 
viable has shrunk dramatically in the last few months!

Attachment: signature.asc
Description: Digital signature

_______________________________________________
Devl mailing list
Devl@freenetproject.org
https://emu.freenetproject.org/cgi-bin/mailman/listinfo/devl

Reply via email to