[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: EXT3 Worries


On Thu, Aug 02, 2001 at 01:03:05PM -0400, Bill Rugolsky Jr. wrote:
> On Thu, Aug 02, 2001 at 05:08:11PM +0100, Stephen C. Tweedie wrote:
> > Umm, there's plenty of precedent.  Rieserfs was an option on SuSE way
> > before it was added to the main kernel.  Raw IO was built into most
> > distributions on 2.2 way before it was released officially with 2.4.
> That is arguably the reason why Linus included it -- because it
> appeared that a number of "real-world" sites were using it in
> production successfully.  Expectations for filesystems are a bit
> different from hardware drivers, etc., and it would not do nobody any
> good if the filesytem were widely deployed and started corrupting
> data.

On the contrary --- ntfs, reiserfs, nfs etc. have all had
data-corrupting bugs when they were first merged in.  ext3 will almost
certainly not be perfect when it gets merged.  But filesystems are
modular: adding a new filesystem which doesn't affect the reliability
of old ones is not a destabilising influence.  Raw IO had several
problems when first merged in, too.

Expectations for new code affecting existing filesystems are very
high.  Expectatations for totally new filesystems are not --- merging
young filesystems is, in fact, often part of the normal process of
getting wider coverage for those filesystems so that they can be
tested more broadly.  That's why we have an "experimental" category
for config options in the kernel.


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]