incubator-bigtop-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From James Page <>
Subject Re: Strategy/Planning Meetup?
Date Thu, 18 Aug 2011 13:47:16 GMT
Hi All

On Thu, 2011-08-11 at 14:11 -0700, Andrew Bayer wrote:
> Given that a bunch of us on the project are in the Bay Area, would
> there be
> interest in actually meeting up in person to start figuring out what
> we want
> to do with Bigtop going forward? Cloudera should be able to host. If
> there
> is interest, it'd be great if we could meet up fairly soon, I think. 

As I won't be able to make the meeting later today here are a few of my
thoughts on what I would like to see from Bigtop going forwards;
apologies in advance if this is a little Ubuntu centric...

1) Standardized set of reference platforms

So at the moment the website states 'Ubuntu 10.10, CentOS 5 and openSUSE
11.4'.  I can't speak authoritatively for CentOS or openSUSE but from an
Ubuntu perspective having reference packages from the Bigtop suite for
the following pattern of releases makes sense:

        a) Current and Previous Long Term Support releases - currently
        Hardy and Lucid; this allows folks using bigtop on these
        releases of Ubuntu (every two years) to have a good support and
        upgrade story.
        b) Current and Previous Interim/LTS releases - currently Natty
        and Maverick - but should shift to Oneiric and Natty once the
        next Ubuntu release is out of the door. Most people who don't
        use LTS releases tend to upgrade more regularly so maintaining
        more than that does not make sense - so having a good upgrade
        path is important.

Obviously where this overlaps with an LTS we can reduce the number of
reference platforms.

Adopting such a policy may mean we need to consider how packaging is
maintained for different releases of a platform.

2) Alternative architectures

ARM on server is getting alot of attention ATM even though 'real'
hardware is not currently in production - this is probably just one
example but maybe we should give consideration to testing Bigtop on non-
x86 architectures as well.

3) Published Packages

For the reference platforms and architectures that we build for/test
should we publish the resulting package artefacts officially?

Anyway - that was just a few of the things I had swilling around in my

Hope that todays meeting is productive (and that I can join the next



James Page
Software Engineer, Ubuntu Server Team

View raw message