On 15/12/13 01:47, Tom Sparks wrote:
>> On Sunday, 15 December 2013 12:18 PM, Ximin Luo <[email protected]> wrote:
> <snip>
>> Hello! Yes, I will be looking into this.
Our basic objectives were to make the node do as much of the work as
possible, so that it's easy to write a plugin, while still allowing a
wide range of plugin implementations. So the API was pretty minimal,
with the intention that registration-time flags and multiple interfaces
could allow plugin authors to implement e.g. flow control if they want to.
> <snip>
>> Of course, Freenet is datagram-based. However, in the long term, Tor also 
>> wants
>> to support a datagram-based underlying protocol, at which point they will 
>> need
>> PTs to also support datagram-based traffic.
>>
> DTN is not datagram-based but bundle based (think of mine email /  epub)
If you can get funding for sorting out transport plugins that'd be
great. There's lots to do, there's a range of possibilities from simple
fake packet headers and streams through to emulating full protocols,
encapsulating data in part of a real stream,
transport-determined/steganographic timing/bandwidth (including
opportunistic data transfer to make efficient use of this), and so on.
Some of this is also required for DTN.

There are plans for DTN but it's a rather longer term project IMHO. It
will require some fairly radical design work, notably full pub/sub
support and probably some sort of long term requests; it may require a
new database, better darknet support etc. Some of that will require
finishing some important core changes. Also, there are different kinds
of DTN:
- I have a slow but realtime connection, and I can also exchange large
amounts of data once a day via USB keys / phone rendezvous. (Easiest to
support)
- I have an intermittent realtime connection (e.g. because of
steganographic requirements).
- I can *only* exchange data via USB keys.

For the last case, I am not necessarily opposed to implementing it in
Freenet, but there are important theoretical issues to deal with e.g.
how to assign locations, and the tradeoffs may be different to those
that Freenet assumes. Generally the view is that high-latency means
publish/subscribe or non-scalable broadcast based architectures (like
Haggle). I'm not entirely convinced of that; IMHO long-term requests are
possible, and may be useful (though pub/sub is of course crucial). But I
haven't looked into what other people are doing in depth.

Haggle implements this in an "opennet" manner, i.e. equivalent to asking
around on a bus who has a copy of a particular subversive document. IMHO
Freenet should stick to darknet. Clearly nobody is going to fund us to
reimplement Haggle (Serval/Rhizome is also worth looking into, it looks
like their site is down at the moment), but we may be able to build
something which is more appropriate for some use cases.
> The best outcome would be if we can unify Freenet transport plugins and Tor
> pluggable transports, so that both projects can benefit from the shared
> research that has already been done.
Definitely worth investigating, there is probably stuff we can do
together in spite of e.g. language barriers. E.g. maybe a common API for
external plugins; most plugins probably won't be written in Java.
> :)
> <snip>
>> Tom, it would be very very useful if you could tell me how much funding you
>> have available to sink into this effort, and any logistics / paperwork-type
>> information. You can send this to me by private e-mail if you so wish.
> I do not have the funds yet,
> as I am in research and pre-proof-of-concept stage
> i'll hopefully be getting the funds next year when I go into proof-of-concept 
> stage :)
>> Ximin

Attachment: signature.asc
Description: OpenPGP digital signature

_______________________________________________
Devl mailing list
[email protected]
https://emu.freenetproject.org/cgi-bin/mailman/listinfo/devl

Reply via email to