Bill Strahm's Weblog

Monday March 07, 2005

IPoIB WG Agenda Slides
Well, this is the fastest way of getting the slides out from this afternoon.

Thank you
IETF Agenda Slides
(2005-03-07 15:26:32.0) Permalink Comments [9]

Friday August 20, 2004

First attempt to post IETF Presentations I want to see how well this works before I go and update all 4 presentations to this blog. To see Hal's Presentation on IPoIB Implementation experiences of the OpenIB group. Please check out the pdf here Implementation Experiences

When I know that this is working - I will publish the other 3 presentations.

(2004-08-20 15:45:00.0) Permalink

Saturday August 07, 2004

Running IP over Infiniband Well, for those that don't know me - I am the co-chair of the IETF's IP over Infiniband working group. This is an interesting pair. IP is the industry standard way to do layer 4 networking (OK - throw TCP on top of it for layer 4). Infiniband is a hardware accellerated Layer 4 protocol suite.

So now we want to run a layer 3 protocol on top of Layer 4 infiniband. Plus there are some issues around not having a broadcast address in Infiniband so we have to create a multicast group in Infiniband to "broadcast" ARP traffic over. There are also interesting issues involving what is a Layer 2 address, and how to turn a long lived Infiniband Layer 2 address (GID, QPN) into something that you can put into an infiniband packet (LID)

How did the working group meeting go. Pretty well. I think we will have the final Encapsulation Draft into the IESG's hands by the end of the month - and make progress on DHCP (need to work with the DHCP WG chairs to have someone look it over their) and the SNMP MIB drafts. I will make it a personal goal to get the Encapsulation and DHCP drafts into the IESGs hands by the November IETF, and the first 3 MIB drafts ready for the IESG as well.

(2004-08-07 13:40:00.0) Permalink

The IETF in action Working in several standards bodies - I have discovered how unique the IETF actually is. I believe that it starts with their unique membership requirements. To belong to the IETF just takes joining the appropriate mailing list and participating. This is unique in that most standards bodies require members to pay large (sometimes into the 5 figure annually) membership dues. This means that effectively only companies can join - allowing their workers to participate.

The other unique feature is how decisions are made. For good or bad - decisions in the IETF are consensus based. Each member (remember anyone can join) has a say into how the standard ends up. Rather than other standards bodies where there are specified ballots (each member - read company) gets a vote - each vote counts as one, whoever has the most votes wins the election. In the IETF there is consensus that is determined by a Working Group chair. This means that everyone can be for something, but a strong minority dissenting view gets listened too, and addressed. This can be good if you have a single expert on a technology that should be listened too - it is bad when someone has an agenda to slow the progress of a working group.

I would like to talk some more in the coming days on the good and bad trends in the IETF, and especially some of the ramifications of their membership model on the future of the IETF.

(2004-08-07 13:23:36.0) Permalink