2002 Linux Symposium

bird of a feather sessions


Note: Additional BOFS will be scheduled on the fly during the event and posted on a noticeboard at the entrance to the event.

Bird of a Feather Sessions (BOFS)

LinuxChix: Women in Open Source

Val Henson

A thorough discussion of women in the computing field and the issues that surround them.

Open Source & Copyright Reform

Russell McOrmond, Flora

In June, 2001, the Canadian Government started a process to reform Canada's copyright laws.| As a response, the Open Source online community set up a forum called the "Canada DMCA Opponents" forum to try to discourage specific policies, such as legal protection for Technological Protection Measures, which have caused considerable problems in the USA with their 1998 Digital Millennium Copyright Act (DMCA).

There are now discussions of starting a more formal Canadian group to continue for more positive changes to laws affecting Open Source and Free Software. The Canadian opposition to a DMCA-like law in Canada has morphed into a more proactive discussion of related laws though articles and forums at digital-copyright.ca. A related organization called "Canadian Open-source Education and Research" (CanOpenER.ca) is also forming to meet the demand for a group that will deal with more than simply copyright law.

Carrier Grade Linux Working Group

Chair Unknown

The OSDL Carrier Grade Linux Working Group was launched in January of this year. The group will highlight its achievements to date, discuss the features of the Carrier Grade Linux roadmap and share the views and opinions of the audience on how to make a Linux-based system to achieve true carrier grade availability.

Bringing Linux into the Data Centre

Chair Unknown

This BoF session will discuss the OSDL Data Centre Linux Working group and its mission to create a roadmap for bringing true data center functionality to Linux. OSDL leaders will discuss what requirements the Data Centre Linux Working Group will be focusing on, features of the roadmap and how to get involved in the Working Group. The BoF will also be focusing on what features, etc. are important to the audience as Linux matures into mission-critical data-centre capabilities.

Future directions of Linux firewalling

Harald Welte, netfilter core team & Astaro AG

The Linux 2.4.x series provided a fundamental redesign of the packet filtering and NAT framework, called netfilter/iptables. This flexible and modular framwork still had it's limitations. This BOF will discuss the recent and upcoming changes during the 2.4.x kernel series, as well as planned and partially implemented changes/extensions for the 2.5.x kernel series.

Incomplete list of topics covered:

  • 2.4.x stuff:
    • The newnat API; supporting connection tracking and NAT for complex protocols like H.323
    • Accessing connection tracking table entries from userspace: ctnetlink
    • Packet filtering and even NAT on a bridge
  • 2.5.x stuff:
    • libiptables: Providing a flexible and extensible API towards all iptables features
    • pkttables: Creating a layer-3-protocol independent layer for rule tables; unifying iptables, ip6tables and arptables.
    • nfnetlink: Move all netfilter/iptables related kernel/userspace communication towards netlink

Linux video development

Billy Biggs

The video BOF will discuss problems and possible solutions to video playback and capture issues under Linux. Discussed will be proposed updates to the XVideo API for interlaced content, a kernel API for refresh sync, discussion on a userspace video4linux library, and quality issues in video drivers.

The BOF will also host a demonstration of some of the proposed updates and why they are necessary. We'll show errors relating to refresh rate sync and deinterlacing. The target audience is kernel, X, multimedia and video application developers.

Linux PPC

Paul Mackerras, Samba

This BOF is an opportunity for developers working on Linux on PowerPC platforms, and anyone else who is interested, to get together and discuss the current state of Linux on PowerPC and directions for future development.

GOSLINGS: "Getting Open Source and Linux INto GovernmentS"

Joseph Potvin, Architect of the first GNU-GPL software release from the Government of Canada

GOSLING is an open community of people who act to ensure that open source methods and software are included in government operations to:

  • Advance the implementation of federated architecture goals, process, standards and methodologies among departments, agencies, and their partners.
  • Facilitate the customization of information technology solutions to business requirements.
  • Improve capacity for information technology quality assurance and due diligence.
  • Serve as a catalyst for practical knowledge-sharing, organizational learning and innovation via public-private-education-civil sector IT collaborations.
  • Reduce costs, increase value-for-money, and achieve faster delivery times for certain government software investments.
  • Generate a broader distribution of benefits from public sector IT investments.
  • Help to level the playing field for companies of all sizes in the government procurement of IT development, customization, integration, support, maintenance, and administration services, by reducing the advantages of prior market dominance.
  • Increase the security of certain information technology applications in government operations.
  • Promote respect for intellectual property.

    Participation in GOSLING involves individuals in their personal capacities, such that content is driven by discussion of research, interests and views of the authors, not the organizations in which they may work.

    This inaugural meeting will be a structured brainstorming session to identify the status, issues, opportunities and constraints related to getting open source and Linux into government operations at all levels, worldwide.


    Robert J. Woodruff, Intel

    This BOF will provide an opportunity for people interested in InfiniBand to discuss architecture, design, and implementation of an InfiniBand stack for Linux.

    InfiniBand is an industry standard that defines a new input-output subsystem designed to connect processor nodes and I/O nodes to form a system area network. The technology is primarily targeted at servers in the enterprise data center. InfiniBand supports storage, networking and inter-process communication (IPC) protocols over one channel based interface.

    InfiniBand is an industry standard that defines a new input-output subsystem designed to connect processor nodes and I/O nodes to form a system area network. The technology is primarily targeted at servers in the enterprise data center to become the system area network within the data center. InfiniBand supports storage, networking and inter-process communication (IPC) protocols over one channel based interface.

    Open source Bioinformatics

    Peter D. St. Onge, SEUL.org

    This BOF will be an opportunity for those currently involved with any aspect of biological informatics to meet and discuss.  Although "bioinformatics" is usually associated with research fields related to gene or gene product (protein) identification and modelling, it has far wider application to other areas of research.

    As researchers, users, programmers and administrators, we are often faced with numerous research, resource and technical challenges on a daily basis. The frequently-iterating, free and public nature of Linux and open source / free software makes them ideal tools for research and teaching. This session will hopefully provide wider perspectives on both the research and technical aspects of biological informatics, and provide a way to highlight current impediments to the use of these tools on a wider scale than at present.

    Linux support for EFI (Extensible Firmware Interface) on IA-64 and IA-32 platforms

    Stephane Eranian, HP

    • a little introduction on what is EFI
    • explain how to develop EFI applications with Linux
    • describe the EFI Linux boot loader elilo
    • describe the kernel support for EFI on IA-64
    • describe the EFI support on IA-32
    • open discussion on:
      • EFI support for IA-32 in 2.5
      • EFI bytecode Virtual machine support in 2.5

    Ext2/Ext3 Development

    Andreas Dilger, ClusterFS.com

    The ext2/ext3 development BOF will be a meeting of developers and interested parties for the future of ext2 and ext3. We will be continuing the discussions started on the ext2-devel mailing list regarding future directions and features for ext2 and ext3.

    Topics will include large directory support, extended attributes, dynamic inode tables, large (> 16TB) filesystem support and other topics of interest. It is expected that most of the primary ext2 and ext3 developers will be in attendance.

    2.5 Kernel Block I/O request splitting

    Mark Peloquin

    In the 2.5 kernel the block device layer has been changed from buffer heads to block io requests called BIOs. The max IO size currently possible today is 64KB, however BIOs are designed to be able to deliver upto a current max of 256 pages. It will be likely be some time before Linux is capable to fully exploit a request of this size. However, IO sizes greater than page size will soon be common.

    Large IO requests present problems for some block devices. Some block devices, based on their internal characteristics, present logical views of storage that are not physically contiguous on disk. Some block devices are capable of utilizing physical areas as small as 1KB. With IO requests exceeding a page size, it becomes possible for a single IO request to spans more than physically contiguous area on one or more disks. In these cases, a large IO request would have to be split or broken up into smaller pieces that effectively "fit" into the physical contiguous areas of the disks.

    The goal of this BOF will be to discuss the various proposed implementations for constructing large IO along with when and how they should be split up. Discussion will cover the pro and cons of these techniques, refined existing ideas, gather new ideas, and hopefully arrive at the best solution(s) for Linux.

    Kernel Janitor BOF

    William Lee Irwin III

    This BOF session will discuss ongoing, planned, and needed kernel janitor tasks, hopefully serve to provide exposure to kernel janitoring issues across a broader cross-section of the kernel for those already involved in janitorial tasks, and also provide a way of communicating what work is in-progress and is needed.

    Kernel janitoring is an important aspect of kernel programming, especially as interfaces and maintainers change, so that both style conformance for increasing maintainability and updates to properly interact with changed surroundings are essential for working code.

    Kernel janitoring will only become more important as the Linux kernel's codebase increases in size and core interfaces change in 2.5. Various strategies for dealing with these issues will be discussed and actions will be coordinated at this BOF session.

    Reliability, Availability and Serviceability issues for Linux (in 2 parts)

    Richard J Moore, TSM RAS Project Lead - Linux Technology Centre

    The RAS BOFs will be an opportunity for those actively working on Reliability, Availability and Serviceability issues for Linux to meet and discuss. RAS had a problem in that it is invasive and not particularly sexy. However, when it comes to providing a platform suitable for running misson-critical Enterprise applications or a Telecommunications Blade Server then RAS becomes a mandatory requisite of the operating system. If Linux is to be successful in these environments it must have system RAS. For that to happen it is essential that the (RAS) Community speaks with one voice and that the various RAS initiatives currently active correctly address the needs of the wider Linux Community as well as those of the Development Community.

    One of our prime aims as a RAS Development Community must be to get core RAS technologies accepted in to the 2.5 kernel. Because RAS affects other system components such as drivers and the kernel itself we also need to agree on a scheme from RAS instrumentation such that it is attractive for the general developer to use.  During this meeting of OLS we are running two RAS BOFs of 90 minutes to enable sufficient discussion time to be devoted to each topic; the agenda for each session is as follows:

    Session 1:

  • Aims, Direction, Adoption Strategy - Richard Moore (IBM)
  • Flexible Crash Dump - status and sub-projects - Matt Robinson(LKCD)
  • Multi-threaded core dump - IBM/Intel
  • Centralised Oops collections (mini system dump) - Rusty Russell.
  • Flexible System Trace - Richard Moore/Karim Yaghmour(LTT)/Hitachi

    Session 2:

  • Event logging - Larry Kessler (IBM) chair
  • Driver & kernel hardening - Lori Matassa (Intel) chair
  • Instrumentation schemes
  • Driver Statistics and Performance metrics
  • Wrap-up - follow-on etc..

    Linux memory managment performance on NUMA machines

    Martin J. Bligh, IBM Linux Technology Center, Scalability group.

    This BOF session will incorporate discussion and brainstorming of what methods and patches will help the performance of the Linux memory management subsystem on NUMA machines.  Recent work has shown that although Linux has basic NUMA memory management functionality, it's performance is far from optimal. Though some of the NUMA performance problems are shared with those of large SMP machines, the non-uniform nature of NUMA memory introduces another whole class of problems and solutions in addition to those of SMP.

    NUMA will play an increasingly important role for future large machines, as the number of CPUs increases and technologies such as Hyperthreading are introduced. Different approaches may be required for machines that have high remote:local memory access latency ratios between nodes vs those that have low latency ratios.

  • 2004
    © 1999-2002 Linux Symposium.  All Rights Reserved.