[Gllug] Nightmare

[home at alexhudson.com]

On Wed, Sep 12, 2001 at 10:28:48AM +0100, tet at accucard.com wrote:
> By the same logic, there were two kernel oopsing bugs found in
> reiserfs last week.

True. Different danger level, though, from what I understood (more difficult
to make happen, I mean ;).

> yet. My objection mostly comes down to the attitudes of those
> involved. I trust Stephen Tweedie, and I don't trust Hans Reiser. Look
> at any LKML archive if you want to know why.

I understand this point, to some extent. I don't think Hans Reiser does
himself many favours.

> The *impression* I get is that ext3 has been well designed from the start,
> whereas reiserfs is a good idea that hasn't been properly thought out
> before they started coding.

I wouldn't say that. I think Reiserfs is pretty decent code - although,
perhaps a little x86 centric. Not that that bothers me any. ext3 is much
less ambitious though, so I would expect it to succeed. It's also building
on top of ext2, which is good or bad depending on your point of view.

> >JFS and XFS I would trust even less - I wouldn't be surprised if they
> >have large bugs wriggling around inside them.
> 
> True, but remember both of those have come from mature codebases. Yes,
> the port to Linux will have involved changes which will have inevitably
> caused bugs, but I'd still feel more comfortable with either of those
> than I would with reiserfs. Just MHO.

>From looking at the patches, they reach in and touch so many core pieces of
code, I'm much more comfortable with reiser. XFS and JFS seem much less
self-contained that the other FSes, or rely on new functionality not
available in standard kernels. I don't doubt the developers, but Reiser has
had a lot of real-world testing. SuSE have been using it as their default fs
for a number of years now. None of the others can claim that yet. 

Cheers,

Alex.

-- 

-- 
Gllug mailing list  -  Gllug at linux.co.uk
http://list.ftech.net/mailman/listinfo/gllug