[cgl_discussion] POC 5/1 conference call minutes

Steven Dake sdake at mvista.com
Thu May 1 09:33:34 PDT 2003


Attendes
--------
Rusty Lynch - Intel
Randy Dunlap - OSDL
Mark Haverkamp - OSDL
Venkat Jagana - IBM
Khalid Aziz - HP
Steven Dake - MontaVista

* Discussion about project tracking matrix.
Talked about Mika's issues.
Talked about past process of maintaining document.  Features were
assigned to owners but not all features assigned.  Not enough people
to track projects.

Have distros not assign lots of engineers, but instead debate what
implementation is acceptable to distros and generate roadmap of when
features will be delivered to their products.  Track problems required
to be solved before integration of projects in distros.

OSDL wants to know what open source projects to assign engineers so
they don't waste energy on open source projects that will never be
accepted.

Suggestoin of distro companies to commit up front what features they
will support.  Purpose is to focus development in OSDL.
Problems are that companies may expose business plan.  Some discussion
about distros not exposing 6 month+ plans because that information
may be private to the company.

Try to make OSDL be clearing house for feature development to
coordinate development between distros, OSDL engineers, OSDL
member companies.

Try to get project maintainers to send status reports for projects
member companies start.  For open source projects that OSDL member
companies not involved in, try to track projects.

Try to get member companies to sign onto projects and track projects
that are open source projects.  Commitment from member companies to
track projects not 100%, need more involvement from member companies
on tracking/committting to projects.

Add field to project list to track what distros need in projects
before they would add the project to their distro.  Describe how much
energy is required to deliver those additions in engineering time.
Describe strategy for enhancing projects

* Project TAHI
Port made of test cases to run on Linux.  Patch sent to project TAHI.
Delay in patch being accepted.

One viewpoint:
Have not released patch on building on multiple OSes because
low priority.  TAHI making changes in build system and they won't
add Linux build support until their build system is reworked.

Different viewpoint:
May not want to include Linux code in their project.  Linux may be
lowest priority or not priority at all.

One solution try to get LTP to integrate TAHI project.

AR Rusty to contact TAHI Linux patch author and ask him to contact
   LTP about including Linux support for TAHI in the LTP.

* OpenHPI
Project moving along rather quickly.  Prototype code working a few
weeks on.  Fundamental questions asked about prototype code which may
require a rewrite.  Rusty now part of SAF.

OpenHPI project not going to create implementation of AIS.

* IPv6
Good news about new functionality added to IPV6.  Multicast listener
discovery support added to linux 2.5.68.  Maintainer has asked for
backport to 2.4 and patches have been submitted to 2.4 and potentially
available in 2.4.21pre.

DHCP v6 conforming to IPv6 spec by June (targeted).

SUSE now contains IPv6 features including ipsec and mobile ip integrated
into CGE service pack from SuSE.

2.5 submission of mobile IPv6.  Maintainer wants ipv4 support added
before ipv6.  Design proposal accepted by maintainer.  Changes being
made for acceptance into mainline kernel.

ICMP MIB support merged into usagi project.  Other IPV6 MIBs are a bit
confusing.

* HDI
Some discussion about MontaVista's project competition with udev.
Major differences are queue of events instead of relying on
/sbin/hotplug, backing data store for device data, state machine
designed to support next-gen hardware designs, and policy mechanism
that allows for creation of combined devices (such as RAID, LVM, etc),
and tracking of permissions and ownership of device nodes through
reboots.

* TIPC
Work ongoing for driver in 2.5 environment.  Cleaning code up to match
coding style of linux kernel.  Trying to simplify code.

* AEM
OSDL comments:
Good idea, but implementation will not fly in mainstream kernel.  Too
invasive into scheduler.

Recommendation is to keep project and try to improve from member
companies as possible.

Need input from distributors to determine if OSDL should apply
resources to work on AEM project.  What is needed to make it acceptable
to maintainers?






More information about the cgl_discussion mailing list