Roadmap

The set of procedures being used to achieve each subsequent release of eBrainpool

What we learnt at Camp - Part 3 Trust

Submitted by jeetu on

 

In this concluding post on stuff we learnt during the eBrainPool workshop at the Chaos Communications Camp (CCC) 2011, I shall focus on discussions we had with regards trust models.

It is increasingly being felt that reputation is the sincerest currency. The social standing of an individual and the positive connections he has in his network i.e. his "reputation" will increasingly determine his "credit worthiness" as perceived by his peers. This is the basic premise of currency models such as Ripple.

One of the biggest challenges and questions posed to us by people with regards eBrainPool is that of security and privacy. In an untrusted group, how do we ensure that people who want to use the software from someone else can trust the  validity of the software? How can both the people in turn also be protected with regards their data - neither party should have access to the others data?

Discussions with people at the camp who have more experience in this field strengthen our belief that the answer to these questions lie in part with technology  /design and to a larger part lie towards the basic tenet of trust. Technology can go only a certain degree but just as in current business and other real world social decisions, it all boils down to the trust between the people engaged in a transaction.

We were privileged to be in the presence with a few people who have worked in this realm and they offered to share their experience and work in trust models. 

Guus Sliepen made mention of his light-weight cryptographic authorisation framework found here. As I understand it this proof of concept library makes use of social connections to further strengthen authorisation.

Caori who has previously worked with Ripple also pointed out his own work in this field. He has an elgg library that can be found here.

In time we hope to have more from both Guus and Caori to speak about these libraries and to hear some more of their thoughts on this and hopefully help in the security and privacy concerns of a technology such as eBrainPool. 

Category: 

What we learnt at Camp - Part 2 Discovery Services

Submitted by jeetu on

 

Discovery Services and breaking dependence from the Olsr plugin

Currently the eBrainPool client uses the Olsr routing messages itself for discovery. Our custom Olsr plugin sends out custom messages that signify the presence of an eBrainPool client on that node. This plugin also scans incoming messages for these custom messages and reports to the client upon discovery of other nodes with the client running.

This is bad design as was pointed out by Mitar earlier and reiterated by others at the workshop. This makes our application unnecessarily dependent on Olsr and also probably puts an extra load in the routing messages being sent out. Ideally discovery should be an independent function and our application should be able to work on any underlying network.

A few options were suggested :

- Zeroconf/Bonjour/Avahi
- Multicast DNS (mDNS)
- Distributed Hash Tables (DHT)
- Telehash

While Zeroconf/Bonjour/Avahi/mDNS work well for smaller and local networks. Distributed Hash Tables have proven successful over the Internet. Telehash works with a routing system based on the Kademlia DHT.

Ideally we need a technology that works well across networks of all sizes and scales from a local network to the Internet. I have to say that I can't wait to start work on this :)

Special thanks to Mitar,Saverio,Pau,Gioacchino,Bruno and Magnus amongst others at the camp for the above suggestions :)

Category: 

What we learnt at Camp - Part 1 Throughput Deterioration in a Mesh

Submitted by jeetu on

 

We've been very fortunate and honoured to have had the chance to interact with some really interesting people at the CCC 2011 and learn from them. We have had many questions with regards mesh, security, privacy and the best way forward and we thank everyone who has attended the workshop and helped show us light as to the best way forward.

This is the first of a series of posts that represent excerpts of a few of those teachings that we hope to put effort towards in the coming time.

Throughput deterioration over multiple hops in a mesh network

One of the things that we were quite eager to determine was how throughput is affected from hop to hop in a mesh network. This is improtant to determine if we run a software from a host that is many hops away how will the usability of the software get affected as the number of hops increases.

We were joined at the workshop by Saverio (part of Ninux) and Pau (http://qmp.cat/projects/qmp). They were kind enough to cite tests done at Battlemesh and their personal experiences that in an ad-hoc wireless mesh network throughput was nearly halved with each hop till about 8 hops down whereupon throughput stabilized.

We demonstrated and tested eBrainPool between two devices (Erle's netbook and my laptop) over wifi and the performance between these two devices which were in radio range of each other was satisfactory.

Pau was kind enough to work with us in trying to get the eBrainPool code onto his laptop so we could have a more detailed series of tests especially in how deterioration affects usability, however we ran into some issues with getting Olsr up and running in ad-hoc mode on his laptop.

Hopefully we will work with the community in trying this out in larger networks and we will have some results to report soon :)

 

Category: 

The Road ahead becomes clearer

Submitted by erle on

For those following the fun and games going on here @ebrainpool, Both Jeetu and myself are currently in heavy coding mode, working on little pieces for the next release.

A quick update on things ... we wrangled back and forth about the timelines that would be required to develop and integrate the authentication and X Forwarding requirements. The good news is we are still sticking with libssh over trying to build a wrapper to handle the default ssh in most systems. Yes we did think about the fact that openssh might be updated more regularly on some systems, whilst keep an eye on security, but we felt the pain from the overall usability in this form was way too high. The current code relies on ugly calls to bash scripts to invoke all this and we really feel the need to do this better.

As of now Jeetu is handling the server side code and I am taking care of the client side code. Key based Authentication and X-forwarding are the key aspects both of us need to build into ebrainpool. Progress has been good, parts of this are working and we are still working towards a release somewhere towards the end of March, or more realistically early April, but we are keeping end March as our target.

In Brief:
1) Use libssh for inbuilt key authentication
2) Use libssh to handle X-Forwarding 
--the above two eliminate the need to rely on the openssh daemon running

2.5) Give some thought towards the cipher used and the option for user to change the default --Not critical right now.

3) Interface to handle storing keys and no manual copy of keys into the directory
4) use a storage engine ? SQlite --needs further discussion

with #1 and #2 above being completed, the code is much easier for people to setup and use, consider a release or wait for #3 above

What after this?
proposed targets after this, #4 and head on towards
5) Fuse (libfuse).. or any other remote FS. Critical here is to handle data privacy and isolation

6) plugin based system to be protocol independent, maybe this should ideally be the next target instead.
7) Avahi, Bonjour for discovery etc... to be seen in conjunction with #6

As always, keep the messages coming and looking forward towards more conversation with all of you.

Category: 

Information on next release (target: March 2011)

Submitted by erle on

Its been a hectic few months. We have been trying to get all this up and running, getting better documentation down, and putting things together to get the development of the next release sorted out. With this new site, coming on, this might be a good time to explain what we have been working towards.
We need to explain the technicals better and that will come online shortly, but as the earlier posts have indicated we use SSH to create the tunnels between the respective devices connected via eBrainPool.  In the current release we need the sshd daemon to be installed on each device and the keys setup in order to be able to connect. We are using scripts to invoke this from the client and we want to change that to a cleaner method.

We are currently working on integrating the ssh system using libssh . Ideally of course this would mean the next release would not need to rely the ssh daemon but would have integrated functionality. This does include an interface to create and edit the keys we use for authentication. 

Would like to complete both these tasks

1) Integration of libssh

2) Interface for handling the ssh settings (remove need for command line settings)

If things take too long we might release 1) and the follow up with a minor release with 2) done above.

More updates on longer term roadmaps to follow.

Category: 

Pages

Subscribe to RSS - Roadmap