ibm-main wrote:
From: "Gerrit Huizenga"
... At the same time, it may be feasible to consider the f series
as a patch set against the -mm tree (e.g. apply 2.6.13 + mm + ckrm
f-series)
as we are running through development, just to keep things a *little*
simpler.
I've been navel-gazing about this.
Is the -mm tree a valid basis for a pre-release "cycle" ???.
As of now, both -mm and mainline are equally convenient as far as
getting the configfs patches are concerned.
-mm has it as part of ocfs2 (though its not independently configurable
from ocfs2) and the mainline patches are conveniently available from
http://oss.oracle.com/projects/ocfs2/files/patches/
for the current linus trees.
But going forward, -mm might be an easier source, particularly if the
attempts to merge configfs with sysfs get any traction. Also, since -mm
is the target tree for the ckrm codebase, it'll be slightly easier to
maintain the ckrm patches against mm directly.
But your points about the instability of -mm are also true...so Gerrit
has to make that call. The comments above are concerned only with the
rcfs/configfs aspect.
No-one in
their right mind is going to get in the ring and argue with Andrew, but this
is (pre-)Alpha in a lot of respects - he's putting the code out there for
people to test. Bits are always being added in and yanked out - in
conglomerates that probably *none* of the respective developers could have
even imagined, let alone tested.
Might we be better off with a continuing CKRM patchset against a mainline
such as 2.6.13/14.
Something like Con publishes as his ck set, but instead
of incorporating our own patches like he does, just incorporate what we need
from other development teams.
Having configfs being reliant on the various cluster proponents agreeing on
a common code-set might be a bit of a worry. Similar (but probably less)
concerns arise with the timely acceptance of cpuset.
Why not have a consolidated set that is on the download page so everyone is
on the same base code - and the project could continue to issue updates
against that as required.
As code gets merged into the mainline, it would (hopefully) be merely a
matter of trimming "our" patchset.
We are exposed to the risk that some of our required code may not make it
into the mainline, but that exposure is no greater, and probably less, than
that from relying solely on -mm. Look at what happened with relayfs.
I know I railed against having to include the configfs patch, but I'm happy
to eat my words on that - a couple of days extra thought and a couple of
beers can do that :o).
Thoughts anyone - or am I just being stupid ???.
Shane ...
-------------------------------------------------------
SF.Net email is sponsored by:
Tame your development challenges with Apache's Geronimo App Server. Download
it for free - -and be entered to win a 42" plasma tv or your very own
Sony(tm)PSP. Click here to play: http://sourceforge.net/geronimo.php
_______________________________________________
ckrm-tech mailing list
https://lists.sourceforge.net/lists/listinfo/ckrm-tech
