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

Re: ReiserFS support

Greg Thomas <ethant_(_at_)_2fortheroad_(_dot_)_net> wrote:
> On Sat, 29 Sep 2001, Fredrik Henbjork wrote:
> > Michael Shalayeff <mickey_(_at_)_lucifier_(_dot_)_net> wrote:
> > > re
> > > how about you take linux stuff off this list to some other,
> > > more appropriate, place.
> > > adding 3 lines at the end do not add to this any more on-topic-ness.
> > 
> > I thought this was a discussion about ReiserFS and journalling
> > filesystems in general, and if it would be a good idea to try
> > to port any of them to OpenBSD. 
> This discussion didn't include any details that developers could use for
> file system development. 

It did include information regarding the real world (in)stability of one
of todays most (over?)hyped file systems. At least I consider that kind
of information to be interesting in an discussion about eventual future
file system development for OpenBSD, since my impression is that there
is a difference between the theoretical world of Research Index and the
real world with real hardware and non-laboratory conditions. Reading
ReiserFS documentation and number of boxen = 1 tests or reviews does
not tell you the whole story about it. As you may know, there is world
outside of the OpenBSD lists, and since *everything* in it is not crap,
it may be a good idea to look for ideas in it about what to do and not
to do in order to further improve our beloved OpenBSD operating system.

> I think the OpenBSD developers know the general benefits of journalling
> filesystems which is all I've seen in this discussion.  

I am certain that the OpenBSD developers know a lot more about file
systems than me. What I had to offer, was some first hand knowledge
about large scale practical use of a particular filesystem, namely
ReiserFS, which some people wanted/discussed porting to OpenBSD.
There are after all not that many situations I know of where you have
several thousand nearly identical boxen running a new open source file
system. That kind of environment, especially when it has severe
stresstesting as part of the daily routine, may help to show _some_
rare bugs in a file system implementation pretty soon, since it is a
lot easier to diagnose the bugs in question thanks to the huge numbers
of similar hardware. The result as I have already mentioned, was that
based on what I have personally seen, heard and experienced, I do not
trust the current implementations of ReiserFS and that my highly
subjective recommendation is that any porting of it to OpenBSD would
not be worth the time it would take. I cannot understand how you can
see this as being totally off-topic in an discussing about ReiserFS
on misc_(_at_)_openbsd_(_dot_)_org?

> And specifically if ReiserFS can't make it as the default filesystem
> on Linux why on earth would other OSes port it?

Well, that was *exactly* what I was trying to say with my original
message... I am sorry if my english is not readable enough for you, but
you can probably blame the swedish public school system for that bug. :-)

Fredrik Henbjork