1 |
On Sunday 17 February 2008, James wrote: |
2 |
> Volker Armin Hemmann <volker.armin.hemmann <at> tu-clausthal.de> |
3 |
writes: |
4 |
> > > > that is bullshit. If you have ever followed the ml you would |
5 |
> > > > now it. |
6 |
> > > |
7 |
> > > It's been languishing in -mm for ages, never mind any progress |
8 |
> > > that namesys itself might make with their own code. |
9 |
> |
10 |
> Well, I'm no Reiser expert, but a few days ago I was reading at |
11 |
> kernel newbies and following some links about the future of the linux |
12 |
> kernel, when I stumbled across something that really makes sense |
13 |
> concerning why many influential kernel devs do not like (trust) |
14 |
> reiser4fs: |
15 |
> |
16 |
> That is the style allows for 'loadable' modules (er the nomenclature |
17 |
> is plugin) and the resulting fear that if reiser4 is 'blessed' and |
18 |
> included into the linux kernel, then those with advanced knowledge |
19 |
> could write very specific modules (of the commercial kind) for niche |
20 |
> feature that just plug into reiser4fs. |
21 |
|
22 |
A further HUGE objection is the extreme difficulty in writing an fsck |
23 |
for such a filesystem. |
24 |
|
25 |
Don't get me wrong, Hans' ideas for reiser4 are extremely |
26 |
forward-thinking and possibly very useful and valuable. Imagine the |
27 |
possibilities - the user could tune the filesystem to do whatever he |
28 |
needed, plug in modules optimized for the data the user is using, even |
29 |
in ways that namesys never predicted. WinFS could actually happen, just |
30 |
not on Windows <evil grin> |
31 |
|
32 |
However, not at the expense of existing deployments and methods. An fsck |
33 |
is an absolute requirement for Linux's largest user-base on something |
34 |
with the scope of reiser4. |
35 |
|
36 |
As you say, the prime reason is probably that other kernel devs are fed |
37 |
up with the person called Hans Reiser and simply ignore him or won't |
38 |
deal with him. That is their right, it applies in every other facet of |
39 |
life. |
40 |
|
41 |
There's a parallel between reiser4/XFS and Con's cpu scheduler versus |
42 |
Ingo's. Linus trusts Ingo and has complete faith that Ingo will |
43 |
continue to maintain his work. He didn't have the same warm fuzzy |
44 |
feeling with Con. There's a truck-load more at stake with new kernel |
45 |
sub-systems that purely code quality, regardless of what the PR line |
46 |
syays |
47 |
|
48 |
|
49 |
-- |
50 |
Alan McKinnon |
51 |
alan dot mckinnon at gmail dot com |
52 |
|
53 |
-- |
54 |
gentoo-user@l.g.o mailing list |