Re: [Help-gnunet] idea!

[Christian Grothoff]

On Wednesday 03 November 2004 17:23, Siim wrote:
> <insert praise here>

Well, thanks :-).

> anyway, i really liked the webproxy, though it isn't really 
> usable. i bet many like anonymous web , so i came up with idea.
>
> freenet's web latency is really low, because of many reasons(busy nodes,
> possibly because big file transactions)... so i thought that there should
> be an web only freenet/gnunet like.. then everybody could use gnunet for
> file transfers and ie gnuweb for web only.... then i though little more and
> though that these both could be integrated... you see even in normal tcp
> internet when there is mayor traffic on the server, accessing websites
> takes very long.. we could blame ISP's for not making traffic policying,
> but that's not right because many big files are transferred through http
> too... so we can leave this out... but with gnunet it's possible to create
> an gnuweb that has very low latency... every node has policy to make web
> transactions first... i mean web transactions have very high priority and
> when you click on the ie mp3 link gnunet takes over... it bet it would make
> many peoples dream come true.

I doubt it. You're forgetting the overheads for doing things in a distributed 
(and for GNUnet at the moment also anonymized) fashion. The result is that 
you really cannot use the system to 'replace' the web for 'normal' operations 
-- the majority of the Internet traffic is never going to be anonymous, the 
cost is just too high to justify that.  Not to mention that in many cases 
anonymity is not desired. 

On your other point, giving high priority to 'small' transfers and low 
priority to 'big' transfers, this will still not give you low latency on 
GNUnet per-se.  At least not anywhere near the what 50-100ms that you get 
with TCP most of the time. Also, what is 'small' and what is 'big' and how do 
you prevent people from cheating?  In particular, a DoS attacker might just 
decide to send out lots of 'small' requests - and the 'big' downloads would 
stall. The current approach in GNUnet where the user that has contributed 
resources gets (proportionally) higher priority is IMO more suited.  Now, the 
current user interface does not allow you to do this, but theoretically it 
would be possible to spend that credit only on the small downloads and let 
the big ones tickle in slowly.  That way the user can decide that he wants 
lower latency for small downloads and really bad service for larger 
downloads, but it would not be a policy enforced by the protocols.  

Anyway, in either case I would not try to sell this as a low-latency service 
comparing it to the ordinary web (or even using that word).  The reason is 
that the costs of decentralization and anonymization are so high that even if 
we had the perfect implementation we will _never_ be as fast or low-latency 
as http while being fully decentralized and/or reasonably anonymous.

> i know it's hard to implement, if it's possible at all(i don't know how
> gnunet works that well), though my knowledge about computers and network
> says that it would be possible... i would gladly help to create it if i had
> proper knowledge about c++(which i will have some day)... but anyway i hope
> you consider this idea of mine... and i hope that i posted to right place
> too.. or should it go to mantis? if it should then could someone tell me...

Ideas and technical discussion (like this) was intended to happen on 
gnunet-developers, bugs or small features go to Mantis, and people that need 
help using GNUnet or that try to understand the system should write to 
help-gnunet. At least that's roughly the idea. 

> why i didn't post this to some freenet forum but here? because i belive
> more in gnunet than freenet, at first i chose freenet because of it's fancy
> homepage, but it seems that gnunet has more in it though it webpage looked
> like some abandoned old projects.. anyway gnunet is surely one fine network
> project with some fine and helpful people doing it :)

Our webpage looks like some old abandoned project!? First of all, which one, 
the one on GNU.org or the one on ovmj.org!?  And second, what made you think 
this? What should be changed? (Note that on gnu.org we follow the GNU-style 
for project webpages.)

Christian