[tahoe-dev] Node correlations - [Was] best practice for wanting to setup multiple tahoe instances on a single node

[Zooko Wilcox-O'Hearn]

On Mon, Jan 16, 2012 at 12:47 PM, Greg Troxel <gdt at ir.bbn.com> wrote:
>
> I'd say write up a plan and send it to the list.

Yay!

> 1) Decide if we are going to trust storage nodes to express the variables that are correlated honestly.  I think it's at least near impossible not to trust them and make progress.

You could have a text file containing each storage server's furl and
its properties in some flat textual format, and put that into your
gateway's node directory (~/.tahoe). Then that gateway would believe
the contents of that file when choosing which shares to upload to
which servers.

As a bonus, this automatically solves the Sybil Attack, in which
someone sets up a large number of storage servers so that by chance
your gateway uses only all or mostly storage servers controlled by
them. Having a locally-controlled file describing storage servers
means that the gateway could have a requirement like "make sure at
least K servers that I upload to appear on this list".

Therefore, I don't think there is any need for the gateway to receive
a description about a storage server directly from that storage server
and then rely on that description. At least as a first cut, it would
seem better to rely on the gateway's system administrator for that.

(I'm sure this will turn out to be inadequate for some uses, but it
seems like a good starting point.)

Regards,

Zooko