[Top] [All Lists]

Re: XFS maintainership (was Re: [XFS updates] XFS development tree branc

To: "Bill O'Donnell" <billodo@xxxxxxx>
Subject: Re: XFS maintainership (was Re: [XFS updates] XFS development tree branch, for-linus, updated. v2.6.28-rc3-9167-g0335cb7)
From: Dave Chinner <david@xxxxxxxxxxxxx>
Date: Wed, 14 Jan 2009 16:50:50 +1100
Cc: Russell Cattelan <cattelan@xxxxxxxxxxx>, xfs@xxxxxxxxxxx
In-reply-to: <20090114054917.GT8071@disturbed>
Mail-followup-to: Bill O'Donnell <billodo@xxxxxxx>, Russell Cattelan <cattelan@xxxxxxxxxxx>, xfs@xxxxxxxxxxx
References: <200901090619.n096Jp20017008@xxxxxxxxxxx> <20090109065935.GA1600@xxxxxxxxxxxxx> <20090109215218.GB10221@xxxxxxx> <4968E61D.6070505@xxxxxxxxxxx> <20090113131855.GB8396@xxxxxxx> <20090114012845.GO8071@disturbed> <20090114054917.GT8071@disturbed>
User-agent: Mutt/1.5.18 (2008-05-17)
[ Resend, this time with an unmangled list address ]

[time to make myself unpopular - the hard questions have to be
asked by someone....]

On Tue, Jan 13, 2009 at 07:18:55AM -0600, Bill O'Donnell wrote:
> On Sat, Jan 10, 2009 at 12:17:01PM -0600, Russell Cattelan wrote:
> | Bill O'Donnell wrote:
> | > On Fri, Jan 09, 2009 at 01:59:35AM -0500, Christoph Hellwig wrote:
> | > | On Fri, Jan 09, 2009 at 12:19:51AM -0600, xfs@xxxxxxxxxxx wrote:
> | > | > commit 0335cb76aa3fa913a2164bc9b669e5aef9d56fa3
> | > | > Author: Lachlan McIlroy <lachlan@xxxxxxxxxxxxxxxxxxxxxxxxx>
> | > | > Date:   Wed Dec 31 12:10:12 2008 +1100
> | > | > 
> | > | >     [XFS] Update maintainers
> | > | >     
> | > | >     New maintainer contact and new tree location.
> | > | >     
> | > | >     Reviewed-by: Bill O`Donnell <billodo@xxxxxxx>
> | > | >     Signed-off-by: Lachlan McIlroy <lachlan@xxxxxxx>
> | > | 
> | > | Folks, it's pretty strange to change the maintainer of a project to
> | > | a person who hasn't done a single commit yet, and without discussion
> | > | with the community who has done almost all of the maintainer work
> | > | for a long time.  I think this should get a little more public
> | > | discussion.
> | It seems fair that SGI would engage in some open dialog with the open
> | source community "that fully embraces XFS",
> | especially give the current levels of "contributions" coming from the
> | likes of Christoph, Dave, Eric etc.
> | Who will have commit access to each tree?
> The SGI designated maintainer (billodo proposed) will have commit 
> access.  Other XFS developers at SGI (as part of the SGI filesystem group)
> may have commit responsibities. These developers include felixb and olaf.

Excellent - three CXFS developers who are already overloaded with
CXFS stuff are going to replace six or seven dedicated XFS engineers
and dedicated XFS QA resources. How is significantly reducing the
resources allocated to XFS going to help SGI fix the problems that
have been occurring in the past 6-9 months?

e.g. How do you does this change help to fix the problems that have
been occurring with patches not being committed to the tree until
months after they were first posted and reviewed by the community?

And how is this going to prevent repeats of the 2.6.27 debacle where
SGI was unable to get all the patches that were queued up out in
time for the .27-rc1 merge window because the dedicated XFS group
were completely overloaded with internal customer problems?

Not to mention the fact that maintainer responsibility involves
*engaging the community*. SGI has pretty much completely failed in
this respect for the last 6-9 months. Any bug that is reported is
generally handled by Christoph, Eric and myself, not anyone from

> | What is the review policy? who has approve/veto power?
> | 
> Community review, with final approve/veto power by SGI XFS group.

So same as before? That's now a fundamentally broken model as it
can result in patches being reviewed internally to SGI and checked in
before community review was performed. e.g. see the recent 64k page size
hacks to dump/restore that got checked in.

> | What about official pull requests, will this always be from who?
> Pull requests will come from the XFS developers at SGI. As of now, they
> include billodo, felixb, olaf.

Pick one, please. I don't want to see Linus get annoyed with
repeated newbie mistakes due to continual rotating of the person
responsible for pushing changes....

Anyway, why should we accept SGI giving commit access only to ppl
who are unknown to anyone outside SGI and have never made any
significant contribution to XFS?

What about Christoph? He seems to be doing all the patch monkey work
right now instead of SGI as well as generating most of the changes
and bug fixes. Why not Eric or myself, who are doing most of the
triage of bugs and review of changes right now?

It seems to me that SGI wants to maintain control without doing any
of the work that having that control requires of them.  i.e. take
without any give....

Point in case: we have a _critical_ 2.6.28 regression w.r.t.
directory handling. The community triaged the bug, the community
fixed the bug and the community reviewed the fix. It got checked
into the SGI controlled dev tree 4 days ago.  Now we are waiting for
SGI to stop playing "let's all be one happy family la-la-la" games
and get off their backsides and *act as responsible maintainers* by
pushing the fix to Linus ASAP.

Please, show us that SGI is really going to act as the maintainer of
XFS. The only thing that will convince me right now that SGI should
continue as XFS maintainer is this:

        "Gesta non verba"



PS: I did say I was going to make myself unpopular :/
Dave Chinner

<Prev in Thread] Current Thread [Next in Thread>