Re: [Help-gnunet] question about namespaces and sporadic updates

[David Roundy]

On Wed, Mar 17, 2004 at 11:47:55PM -0500, Christian Grothoff wrote:
> On Wednesday 17 March 2004 06:07 am, David Roundy wrote:
> > Is there any way to recursively add a directory structure to my
> > namespace?
> 
> Yes, use the options "-rb" (-r for recursive processing, -b for "build
> directories").  That should give you almost exactly what you want.  Oh,
> and I'd suggest using gnunet-gtk for downloading directories, it gives
> you a nice way to "browse" the directory structure (do a namespace search
> for the root, then you can start to navigate directories).

What I was looking for was for a solution that wouldn't require downloading
directories (which are large), since I already know the filepath for the
file I'm looking for.  And in fact, the only reason I chose gnunet to look
into was the fact that it looked like it had decent support for command
line usage.

It seems (just from experiment) that -rb doesn't actually work like I would
think.  It seems that it only adds the contents of the directory to the
"directory" files, but not to the namespace, as far as I can tell.  So to
get the file "foo/bar/baz/hello" you'd need to fetch the directory listings
of "foo" then "foo/bar" then "foo/bar/baz" and then you'd be able to find
the file "foo/bar/baz/hello", but then maybe I just wasn't phrasing my
search right when I tried to locate "foo/bar/baz/hello".

> > It would also be nice to have a flag to tell gnunet-search to exit
> > after it has found one reasonable match (where "reasonable" may be a
> > bit tricky to define).  A timeout of 1 second (and the timeout doesn't
> > seem to accept floating input) would add about four hours to
> > downloading the (rather large) repository I was considering making
> > available via gnunet.
> 
> Eh, what?  I'm not sure I understand what you're talking about here.  How
> can 1s take 4 hours???  I don't see why anyone would want floating-point
> timeouts, but it should be trivial for anyone to add an option "exit
> after n search results".

1s takes four hours when multiplied by 16,000 (and then round up).

> > Perhaps I should just give up on the idea of using gnunet, since it
> > doesn't seem to be designed to scale for the purposes for which I'd be
> > interested in using it, which is as an alternative transport mechanism
> > (that is, alternative to http, scp, ftp etc) for darcs repositories.
> 
> I don't know what darcs is. Also, http/scp/ftp etc. are not really
> anonymous, so there will always be some performance hit right here. If
> you don't need anonymity (or deniability), don't use GNUnet's AFS.  In
> the future we might have non-anonymous (and thus faster) transport
> mechanisms build on top of GNUnet.  Either way, GNUnet makes no sense at
> all as a replacement for ftp or scp, if you need scp, use scp (or with
> updates, rsync / rdist!).  We're not trying to build a replacement for
> these fine tools here.

When I say alternative to http, scp, ftp, etc, it means I already support
all of the above, but was thinking of users who don't have access to a nice
server.  Basically I was thinking of gnunet as a replacement for something
like bittorrent or napster, i.e. a reasonable way for people who don't have
much bandwidth or don't have an entirely stable internet connection to
share files with a large number of people by taking advantage of the
bandwidth of their downloaders.  But I guess perhaps it's not that either.

But seriously, I don't see how you can provide much plausibility to the
deniability if noone uses gnunet for non-subversive purposes.
-- 
David Roundy
http://www.abridgegame.org/darcs