Re: [Help-gnunet] INDIRECTION_TABLE_SIZE and download speed

[Christian Grothoff]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On Sunday 08 September 2002 03:27 am, Tracy R Reed wrote:
> On Sat, Sep 07, 2002 at 04:35:25PM -0500, Christian Grothoff spake thusly:
> > > Wouldn't it be possible to automatically send out queries for random
> > > blocks of a file to ensure that the file is available. If such queries
> >
> > Yes, this idea is feasible (except that in order to download 'random'
> > (leaf) blocks, we'd first have to download a couple of inner blocks
> > (those blocks
>
> Alternatively, couldn't such metadata expire after, say, an hour? That
> would avoid having stale info floating around.

No, that is not possible (you can not tell what the metadata refers to, you 
can not tell if it is stale, etc.). 

> I can currently search for stuff and often see the exact same files that I
> have seen for days but have never been able to even begin downloading.
>
> Also, I have two nodes running. I inserted a bunch of content on one node
> and searched for it on the other node. Both nodes were well connected with
> 10 and 13 hosts connected each. It didn't find it even after a very long
> time. With only about 20 nodes in the network how can this be?

This should not really 'be'. Sounds like a routing problem (aka 'bug') to me. 
What you can do to investigate is print the queries that node A sends out and 
print the queries that node B receives. If the queries never reach node B, 
we'd have to investigate why (B not connected? B connection broken? B drops 
packets? B not directly connected to A and intermediary 'C' fails to route 
properly? B directly connected to A but A does never send query to B? A drops 
outbound queries and they never reach anybody?). Many possibilities here, and 
they may relate to the download speed much more than the stale content and 
other issues that have been brought up in this thread.

> Maybe my nodes are both just terminally poor although I can't understand
> why since I have tons of bandwidth, cpu, and I have inserted a fair amount
> of content on each.

Even if they were very 'poor', they should (with a 1/128 chance in 0.4.6c) 
connect directly to each other and since they're both idle, they should 
answer each other's queries. 

Christian
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.0.6 (GNU/Linux)
Comment: For info see http://www.gnupg.org

iD8DBQE9e7vu9tNtMeXQLkIRAiMiAJ9qurmZ+nDPWpTIyV4EPK+btZCHZgCgqN4w
QbiLjN3IlFTV6502KiEeKME=
=zPNu
-----END PGP SIGNATURE-----