xfs
[Top] [All Lists]

Re: [PATCH, RFC] - remove mounpoint UUID code

To: "Eric Sandeen" <sandeen@xxxxxxxxxxx>
Subject: Re: [PATCH, RFC] - remove mounpoint UUID code
From: nscott@xxxxxxxxxx
Date: Wed, 12 Mar 2008 07:09:56 +1100 (EST)
Cc: "xfs-oss" <xfs@xxxxxxxxxxx>
Importance: Normal
In-reply-to: <47D5DE13.8030902@sandeen.net>
References: <47D20F78.7000103@sandeen.net> <1205196252.15982.69.camel@edge.scott.net.au> <47D5DE13.8030902@sandeen.net>
Sender: xfs-bounce@xxxxxxxxxxx
User-agent: SquirrelMail/1.4.8-4.el4.centos
> Nathan Scott wrote:
>
>
> (hope you didn't too much mind my quoting you in this thread) ;)

(hope you didn't mind too much my dissing your cleanup) :)

>> Since effectively all versions of XFS support this feature ondisk,
>> including complete support in recovery, it would be better IMO to
>> leave it in for someone to implement/experiment with the syscall
>> and auto-mounting userspace support.  That would then require no
>> new feature bits, mkfs/repair changes, etc.  There is effectively
>> zero cost to leaving it there - and non-zero cost in removing it,
>> if our seriously bad regression-via-cleanup history is anything
>> to go by ... :|
>
> the only cost to leaving it is having another instance of "ok now what
> the heck is THIS?!" ... death by a thousand cuts of xfs complexity.  But
> yeah, removing it has some risk too.

So, document it and move on.  It would be a fun little project to go and
experiment with this code a bit.  It amounts to a trivial amount of code
at the end of the day, and theres certainly nothing "complex" about it.

>> It would be really unfortunate to remove this, and then find that
>> it was useful to someone (who didn't know about it at this time).
>> OTOH, if there is definately never ever any chance this can ever
>> be useful, then it should indeed be removed. :)
>
> Well I'm not hung up about it.  If anyone thinks it'll be useful, I'm
> not bothered by leaving it as is.  So, Nathan, what are your plans for
> this code? *grin*

I don't have any immediate plans.  I can imagine it could be used to
stitch parts of the namespace together in a filesystem that supports
multiple devices (in a chunkfs kinda way) ... or maybe more simply
just an in-filesystem auto-mounter.  *shrug*.  But its there, the tools
support it (once again, I didn't see a userspace patch - hohum), so I
would vote for leaving it in its current form so some enterprising,
constructive young coder can try to make something useful from it
at some point.  :)

cheers.




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