Subject: Re: dynamically growing ffs
To: Bill Studenmund <>
From: Julian Assange <>
List: tech-kern
Date: 09/14/1999 10:49:00
> On Tue, 14 Sep 1999, Julian Assange wrote:
> > (my application is actually a distributed network drive, but the
> > principals are the same). Bill, Jason -- you did some ffs work
> > recently, what do you think?
> What are you using to distribute the network disk?

tcp initially. however I'm investigating various udp multicast
solutions for redundancy and speed on local networks.

In writing my cryptographic file system, marutukku, I developed a
fast scatter-gather userland block device. Initially this was simply
to kick all of the crypto-bloat out of the kernel, and ease debugging
of the crypto-code. However I always had the view that one day this
code would become useful for allowing remote network access to
encrypted blocks on another machine via a low-weight
block-request/block-write network protocol. Authentication is a
non-issue as only the cipher text ever passes over the network
(because the blocks are encrypted/decrypted only on the client).
But it seems to me this code is also incredibly useful for rapidly
proto-typing a conventional distributed block device also.

One of the benefits of a distributed block device is that it's very easy
to grow. However if our file-systems can not grow, there is little point.