Other Distros (NOT A FLAMEWAR TOPIC!!)

[Alan McKinnon]

On Sunday 09 April 2006 00:30, Daniel Carrera wrote:
> Alan McKinnon wrote:
> >>I remember the joys of configuring the inode density of my disk
> >>(who the hell needs to configure that?!).
> >
> > Anyone using ext2/ext3?
>
> Only if you plan to have a gazillion files that are no bigger than
> a few Kb. Not the most typical scenario.

I always thought I was an atypical user. Now I have proof:

develop genkernel # find / -type f -print | wc -l
1066146

Not quite a gazillion files, but close :-)

> > That's the main reason I use ReiserFS. The other reason is not
> > having to suffer through fsck every x mounts
>
> ReiserFS doesn't require fsck? I'd like to hear more.

It does fsck, but not in a rigid pattern like ext3 does. At boot time 
the file system is checked and fsck is run if and only if required. 
As I understand it, an fsck only needs to be done if the journal 
indicates that data wasn't committed to disk such as in a power 
outage or other interruption. The data is already in the journal so 
the write can be completed.

fsck'ing every x mounts just in case ext3-style feels wrong to me. If 
there's a possibility of inode corruption I'd say there's serious 
problems with that file system and an fsck is unlikely to find and 
fix everything.

ext2/3 is a good reliable filesystem and gets the job done. But it's 
rigid inode structure that can't be changed without mkfs grates the 
engineer in me. I ran out of inodes once, that was enough to convince 
me that adaptive file systems are worth looking into. LVM has similar 
reasoning behind it, the hassle of changing your partition layout 
when you discover your initial estimates were out is a hassle that 
shouldn't happen IMHO.

And the enhancements being considered for reiser4 are uber-cool :-)

-- 
Alan McKinnon
alan at linuxholdings dot co dot za
+27 82, double three seven, one nine three five