PCFS mount options "noatime"/"atime" [PSARC/2007/415 FastTrack timeout]

[Frank Hofmann]

> From carlsonj at phorcys.east.sun.com Wed Jul 18 16:04:00 2007
> MIME-Version: 1.0
> Content-Transfer-Encoding: quoted-printable
> Date: Wed, 18 Jul 2007 09:55:55 -0400
> From: James Carlson <james.d.carlson at sun.com>
> To: Joerg.Schilling at fokus.fraunhofer.de (Joerg Schilling)
> Cc: opensolaris-arc at opensolaris.org, jk at tools.de
> Subject: Re: PCFS mount options "noatime"/"atime" [PSARC/2007/415 FastTrack
> 	timeout
> X-Fraunhofer-Email-Policy: accepted
> X-Scanned-By: milter-spamc/1.10.376 (mailgw12 [153.96.1.24]); Wed, 18 Jul 2007 16:03:53 +0200
> X-Spam-Status: NO, hits=-2.60 required=5.00
> X-OriginalArrivalTime: 18 Jul 2007 14:04:00.0869 (UTC) FILETIME=[7DC9E150:01C7C944]
>
> Joerg Schilling writes:
>> Jürgen Keil <jk at tools.de> wrote:
>>> Wouldn't it be possible to optimize out writing back the FAT
>>> FS metadata in the "atime" case, when we detect that the
>>> day-granular access timestamp hasn't changed?

This is happening in the new code, indifferent of the atime/noatime mount 
option setting. It's just a reasonable optimization. The answer is yes.

>>
>> This would be possible with a rewrite, but it seems that the people
>> who decide on that are not interested in a rewrite.
>
> The "people who decide on that" are in fact the contributors on
> opensolaris.org.  If you want to change it, then do so.  Don't
> complain about what others may or may not find worthwhile to do with
> their time -- it's not just misleading, but it's actually
> counter-productive.  It makes it sound as though there's some central
> authority blocking progress on pcfs, and that's not the case.
>
> OpenSolaris is still rather new, as open projects go.  It's
> disappointing to see prominent contributors already trying to tear it
> down by inventing conspiracies that don't exist.

I'd rather set up a skunkworks project on OpenSolaris to create a 
from-the-ground-up reimplementation, than continue to try a rewrite of 
PCFS from within Sun, getting the community "involved". Anything, as long 
as it's not called "PCFS".

That's because an open project can make its own rules, only subject to the 
constraints of the communities and finally the OGB. Opening an internal 
product so that the product team decides not only on the "what", but also 
on the "how" (the set of policies that, for the project, governs who is an 
acceptable reviewer, and how reviews and tests are done), that has, for me 
at least, proved to be beyond my abilities.

Just me, though. Not everyone might've made the same experiences.
FrankH.

(I'm not on opensolaris-arc at opensolaris.org, please cc: me if you expect 
me to answer - thx.)