Subject: Re: Why not softdep per default?
To: Karsten Kruse <tecneeq@tecneeq.de>
From: Hauke Fath <hauke@Espresso.Rhein-Neckar.DE>
List: current-users
Date: 03/29/2005 23:54:04
At 23:16 Uhr +0200 29.3.2005, Karsten Kruse wrote:
>> I feel that in case of filesystem code, it's good to err on the safe side.
>
>Agreed. But that does't mean to deny any new stuff just because it's safe
>to keep the old around.

It's not denied in any way - you are free to set the option in sysinst for
any filesystems you create. You are free to add the option to /etc/fstab at
any time later.

It's not _forced_ on you. And that's good, in my book.

>After all i get the impression NetBSD's softdep implementation is shaky
>and few people trust it.

That impression is wrong.

Personally, I've got filesystems mounted softdep on most of my machines -
mac68k, sparc, i386, all running fine.

But there are exceptions where things didn't work that well. The pr 26825
which I hinted at is actually a scsi driver issue, not softdep related. But
the delayed writes to softdep or async mounted filesystems push the machine
over the edge.

>> If I want linux, I know where to find it.
>
>I don't see what that has to do with softdeps.

Linux has a tradition of preferring speed over safety in filesystems. The
Berkeley ffs has a tradition of coming up unharmed out of any sort of crash
(discounting driver bugs). I know where I'd put my money.

	hauke

--
/~\  The ASCII Ribbon Campaign
\ /    No HTML/RTF in email
 X     No Word docs in email
/ \  Respect for open standards