[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [GNUnet-developers] Replacing lookup

From: Tracy R Reed
Subject: Re: [GNUnet-developers] Replacing lookup
Date: Mon, 10 Mar 2003 20:35:52 -0800
User-agent: Mutt/1.2.5i

On Mon, Mar 10, 2003 at 06:54:38PM -0500, Christian Grothoff spake thusly:
> So far, I have to agree. Also we should learn a bit from Freenet here. Their 
> home-grown (!) datastore was troublesome for Freenet as far as I've heard. So 
> maybe the problem is truely bigger than what we thought. And even if we're 
> doing better than Freenet did by using a well-tested library (gdbm/tdb) to 
> start with, I'm happy to consider going to MySQL at this point. MySQL is 
> GPLed, so this is no problem either.

Freenet has had two datastores. The first one used one large file
containing their own custom sort of filesystem. It turned out to be
extraordinarily difficult to debug and corrupted itself often causing the
whole file to have to be erased causing the node to lose its cache. This
really hurt the network. The idea was that it would be more secure to have
everything stuffed into one file and encrypted. They then moved to a
"native" datastore where each object is created as a file in the
filesystem. This worked out MUCH better.

Either gdbm or MySQL is a huge dependency to put on GNUnet which virtually
guarentees that no Windows user will ever run it. As long as we don't care
if the network ever gets very large that may be fine. But Freenet has been
extensively analyzed, simulated, and otherwise mathematically modeled, and
it seems that the more nodes participate the better it will work. Lacking
equivalent analysis I don't know how GNUnet compares but I don't think any
Freenet/GNUnet system will be very successful if it ignores the Windows

Tracy Reed

Attachment: pgpuYqoMQkM2Z.pgp
Description: PGP signature

reply via email to

[Prev in Thread] Current Thread [Next in Thread]