[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[GNUnet-developers] Re: [Help-gnunet] Messaging system planned? (Was: Re
[GNUnet-developers] Re: [Help-gnunet] Messaging system planned? (Was: Re: Etwas Content)
Sat, 22 Jun 2002 16:06:57 +0300 (EEST)
On Fri, 21 Jun 2002, Christian Grothoff wrote:
> > Is there some kind of usenet/frost like messaging system planned? I
> > think it would be very usefull for this kind of messages.
> Well, I would like to see one, but while I'd help with the design and
> questions, nobody has really started to work on it. If I remember right, Igor
> Wronsky was considering writing an RFC as a starting point for a design
I was considering writing the whole app, after some good design
had been agreed upon. To cut the whole story (relatively) short, an
ultimate design wasn't to be trivially found (we had quite long
off-the-list discussion) and up to date, I've experienced
several technical problems related to the current gnunet
implementation. This means, before
a) gnunetd runs without trouble for several days in a run,
- no massive incoming/outgoing udp flooding
- participating usefully in the network
b) content can be relatively reliably inserted and retrieved
(last time I tried, gnunet-download hanged gnunetd <grin>,
which shouldn't happen.)
are achieved, I don't have enough faith in the system to code
a potentially massive application on top of gnunet. ;) But
don't take me wrong, those two points might be realized by
the time of writing of this, I haven't yet tested todays CVS.
As I said to Christian, Frost could be easily ported to
gnunet but I don't think we want that. There is the XML
style, for one thing, and Frost is totally vulnerable to
spamming. If somebody though doesn't agree and takes on
porting it, I too can give hints of what should be modified
to take advantage of the differences between gnunet and
freenet. For example, I gnunet, the incrementing counters
in keys are not needed.
> far, I would use some old-fashioned struct containing some magic number,
> version number, pseudonym, a public key, an RSA signature of the (hash of)
> the rest of the directory, maybe some checksum, maybe (optional?) a
> timestamp, and a list of RootNode structs (as used in GNUnet).
The timestamps was one of the main points we didn't agree upon. I
argued that timestamps are necessary (should even be included in
the KEYWORD, as in frost) so that the app does not unnecessarily
retrieve and populate network with antique information. Christian
argued that the timestamps opened up doors for attacks. I don't
disagree, but I don't see any option either.
More views on this?
ps. It seems David had an emphasis on the *messaging* property
of Frost, whereas we have mostly discussed the file/hash
directory part. Messaging probably needs somewhat different
approach, and I don't know if an adhoc system (like Frost)
should be developed specifically for gnunet, or if e.g. just
a "gnunet nntp daemon" should be done so that any nntp
newsreader could be used to communicate through gnunet.
I certainly have *not* considered the options (different
protocols, messaging systems/standards available), and am
not familiar with them. I have even thought usenet a bit
out of date, but there it is, still...