[CUWiN-Dev] updated roadmap with notes
Brandon Bowersox
brandon at ojctech.com
Tue Jul 11 13:10:40 CDT 2006
Here is an updated roadmap draft based on emails, discussion with
Dave, Ross, Sascha, and others. This gives some idea of where we
plan to go with version numbers and features in the coming months.
There is also a large list of research ideas and features that are
not scheduled for a specific list but may happen any time depending
on people and technology. Ideas and comments are welcome. -Brandon
____CUWiN ROADMAP_______________________________
0.7.0 bug fixes in hsls; multi-architecture build - August 2006
0.8.0 improved web interface/nodeconfig for multi-radio - September 2006
0.9.0 meraki mini support (pending availability)
1.0.0 bandwidth shaping; discover and select gateways (pending SoC
project)
____CUWiN 1.0 REQUIRED FEATURES_________________
- ROBUST AND DE-BUGGED
fix bugs such as internet gateway default route problem
- NODECONFIG / TURNKEY
multi-radio setup so nodes can provide access points
____CUWiN 1.1+ FEATURES__________________________
- Discover and select gateways [Michael Earnhart SoC project 2006;
target for 1.0.0]
When there are 2 gateways and one is crowded or slow, pick the
other. End users will understand and judge CUWiN based on getting
easy fast Internet access. Is this called "gateway-selection protocol"?
- Name Service
Communication and content-sharing among community participants.
The "ad-hoc name service" is 80% complete. Specs are at http://
svn.curireless.net/svn/cuw/trunk/doc/local/ name-service*. Dave has
code for the reflection aspect. We need to do the nodeconfig part.
Need to debug name service bridge. We can also add ability to have a
DNS name on the internet and advertise servers. Name service is overdue.
- WifiDog
Captive portal and authentication. When nodes provide access
points, we want to allow the node administrator to require login from
a captive web page before they get network access. Registration might
just require a MAC address or email.
We would operate a back-end wifidog server; Dan Merideth is
setting this up. Each node would be a client. Bryan, Chase or Bill
Comisky could help import and Dave has instructions: http://
svn.cuwireless.net:8080/svn/cuw/trunk/doc/local/3rd-party-src.d/
general-import-insns.
____COMING FEATURES NOT SCHEDULED FOR A SPECIFIC RELEASE____
- ETT: Another mesh routing metric for experimentation; extract from
kernel and inject into hslsd.
- HSLSD "feature-complete": adapts more quickly to changing networks,
improve bootstrap, handle merges, handle bidirectional links
according to BBN TR 1301. Add linkstate locking and decision/hello/
lsu threads, so real-time performance for flooding/hello protocal
improves.
- Virtual AP: NetBSD team will be porting this from FreeBSD, we hope.
This will allow a node with one radio to be both a backhaul and
access point.
____COMING PLATFORM/DRIVER SUPPORT NOT SCHEDULED FOR A SPECIFIC
RELEASE____
- Linux Port: CUWiN is tightly integrated to BSD so a Linux port will
not be trivial, and will require a significant contribution to
maintain over time. For these reasons Dave feels a Linux port is a
low priority. Jennifer Hou and other supporters desire a Linux port,
however, and it may grow the coder and user base. If someone is
willing to perform the port and maintain it, Dave is ready to accept
this contribution. ETX and HSLS are already ported to Linux. Note:
Linux cross-build already exists and is used by CNT and others.
- There is a large list of cards that need drivers and consumer-grade
devices we want to support. Meraki Mini is the important one that we
hope to provide by 1.0.0.
____RESEARCH IDEAS - NOT DESIGNED OR SCHEDULED FOR ANY RELEASE____
- support nodes sleeping for power savings, expand HSLS by adding
sleeping logic based on WSN/sensor techniques [offer by Ashish in
July 2006]
- IPv4/IPv6 NAT Traversal: Teredo would allow IPv6 services in the
mesh to be available at IPv4 addresses from the Internet. Tunnels,
such as OpenSSH, may be needed for end-users to make services
available. Alternately, assign each computer a global IP address?
- Channel Agility: Research ability to coordinate channel changes or
operate on multiple channels. Dave feels this is a significant
amount of work. This will improve speed and efficiency.
- Remove Quagga to compress CUWiN size. Currently we are only using
Quagga for it's ability to enter routes into the kernel routing table.
- VoIP: R&D is needed to reduce latency. Also test video.
- Bandwidth shaping
- QoS
- Collect and Publicize Node Data: needed by administrators.
- Scalability to clouds of more than 50 nodes.
- Configuration information store.
- VizRoute: improve, switch from GD to another library, release for
net admins
- Low Priority Research: OPN / MS Research's MultiNet. Setting
appropriate carrier-sense thresholds. Transmit power control; self-
interference mitigation such as Least Interference Routing.
Multicast. Act on interface events (up/down/add/delete) and media
events (active/no-carrier). Redistribute routes from other protocols
to hsls.
- Compress CUWiN to 8MB: to move to smaller, cheaper devices.
- Testing Harness Platform: A way to test cuwin routing, hsls, to
identify bugs and performance bottlenecks in a controlled environment.
- 802.11 link adaptation: allow nodes to find best link to gateway
based on packet loss and return statistics. Also need 802.11 link
layer instrumentation.
More information about the CU-Wireless-Dev
mailing list