Re: Patches for proposal 155: Four Improvements of Hidden Service Performance

[Roger Dingledine <arma@xxxxxxx>]

On Wed, Oct 15, 2008 at 10:33:00AM +0200, Karsten Loesing wrote:
> >> Index: /home/karsten/tor/tor-trunk-155-patch2/ChangeLog
> > 
> > Isn't this patch a smaller version of patch4? That is, if we started
> > enough circuits to begin with, can we just skip this step? It seems
> > odd to apply both patch2 and patch4, and it seems like patch4 is the
> > more general.
> 
> Ah no, this patch affects circuit extension to an introduction point on
> client side, whereas patch 4 deals with establishing introduction points
> by the service.

Ah ha. I've clarified the changelog entries then, and committed it.

> >> Index: /home/karsten/tor/tor-trunk-155-patch3/ChangeLog
> 
> All in all, I'd say we hold back this patch for now until we know better
> whether this is a problem at all. It probably wouldn't hurt to add it.
> But it doesn't make sense to add this much new complexity (your proposed
> code change reduces that to some extend, but still) for an uncertain effect.

Ok. We'll skip it for now.

> >> Index: /home/karsten/tor/tor-trunk-155-patch4/ChangeLog
> >> ===================================================================
> >> +    - Start building more server-side introduction circuits than needed
> >> +      (five), pick the first three that succeed, and use the others as
> >> +      general-purpose circuits.
> > 
> > Looks plausible. I'm going to review the patch more tomorrow and then
> > check it (or something like it) in.

I've checked it in, along with a clarifying comment so the next person
to try to figure out this crazy design has a prayer. :)
http://archives.seul.org/or/cvs/Oct-2008/msg00089.html

Thanks!
--Roger