[lsb-discuss] LSB conference call agenda (2011-05-25, 11am ET)
wookey at wookware.org
Wed Jun 1 10:22:17 PDT 2011
+++ Jeff Licquia [2011-05-25 17:59 -0400]:
> On 05/25/2011 04:35 PM, Jon Masters wrote:
> There were a few questions that we had:
> - How committed is everyone to an ARM profile for the LSB? We're
> pretty sure we'll need help getting the architecture right, not to
> mention our own limited resources and the other tasks we have. Can we
> count on Linaro/Debian/Red Hat/whoever people to participate?
I've kicked off some conversation to try and determine that. I think
it will happen, but as there probably aren't any actual users yet it
requires someone to have a bit of vision and decide that it's worth
doing. It looks like SteveMcIntyre is going to take this on within
[weekly call happened whilst I was writing this mail]
> - It was pointed out that, historically, ARM hasn't had a single
> architecture to standardize. I've seen the posts here that indicate
> that this is changing, with Debian, Ubuntu, Fedora, and MeeGo all
> converging on a single architecture. There has been a lot of optimism
> that this will succeed, but it might be good to understand how it might
> not, and why past attempts (if any) did not.
There haven't been past attempts to standardise a single ABI - just a
set of permitted ABI options (the EABI work published around 2005,
which unified ABIs in Linux, WinCE and SymbianOS, but there were still
various options used between and within OSes). That spec is adhered to
in the ABI options chosen for the new ports I've already posted about.
I guess the risks of failure are around insufficiently-specified
things ending up different in different distros like the linker-name
encoded in elf binaries. The Debian and Ubuntu ports are essentially
being done by the same people so compatibility there is pretty-much
guaranteed. Fedora are now plugged in to that process so that should
work out too. I hope the Meego people haven't done anything
incompatible already :-)
Doing the LSB work actually helps ensure that this is done
right because things become checkable and fully-specified. This was
the context in which it came up originally - release manager tpyes
wanting to know how they checked that they binaries thet were
distibuting actually conformed.
> - Is there interest in an ARM profile on the enterprise distro side?
> We are working to broaden the LSB beyond the enterprise distros, and
> hope to do so as part of LSB 5.0, but it's still worth knowing if we can
> expect a RHEL or SLES for ARM at some point, as just two examples.
The Fedora work will almost certainly end up in RHEL eventually. It's
safe to say that you'll be seeing ARM in server rooms in the
not-too-distant. Prototypes are already in existence - people will be
flogging hardware fairly soon.
> - In addition to time and attention, the LSB workgroup would need
> hardware for development and maintenance. Is there anyone willing to
> donate hardware resources for that purpose?
Almost certainly. Debian has porter machines already for Debian
developers, so does Ubuntu (I'm not sure what the requirements for
access there are). I'm sure Linaro can get someone a board to work on
if that's what's needed. (Or you can just buy one for <$150)
So, as mentioned on the call the next steps are to make an 'arm
branch' and database instance accessible to us 'arm people' (and give
us a clue on how to make changes or send in data) and we can start
working through making it correct.
Principal hats: Linaro, Emdebian, Wookware, Balloonboard, ARM
More information about the lsb-discuss