Quotes of the week [LWN.net]

/* As Linux tends to come apart under the stress of time travel, we must be careful */

This many years into the effort we ought to be slicing and dicing volumes as second nature, changing configuration on the fly, transparently expanding, shrinking and migrating filesystems, and many other things that ZFS and GEOM are already doing and we are not. It is not so much that device mapper is incapable of such fancy tricks, but that we have taken a very powerful kernel subsystem and hobbled it with a nearly unusable application interface. Think about a jet turbine racecar with a two inch air intake.

-- Daniel Phillips

to post comments

Quotes of the week

Posted Mar 6, 2008 13:11 UTC (Thu) by davecb (subscriber, #1574) [Link] (3 responses)

A hardware/software codesign engineer of
my acquaintance once said "Hardware engineers
put capabilities into chips. Driver writers 
take them out"

--dave

Quotes of the week

Posted Mar 6, 2008 13:17 UTC (Thu) by BenHutchings (subscriber, #37955) [Link] (2 responses)

This also works if you s/capabilities/bugs/.

Quotes of the week

Posted Mar 6, 2008 15:15 UTC (Thu) by AJWM (guest, #15888) [Link] (1 responses)

Bugs are just capabilities you haven't yet figured out a use for.  ;-)

Quotes of the week

Posted Mar 6, 2008 17:55 UTC (Thu) by jospoortvliet (guest, #33164) [Link]

Hehe, just like some programmers at my job often tell me: "it's a feature, 
not a bug" ;-)

Quotes of the week

Posted Mar 6, 2008 15:19 UTC (Thu) by incase (guest, #37115) [Link] (1 responses)

Note that the sentence quoted to be by Rusty is from inside a diff and that the sentence was
there even before the changes, just with a different introductory sentence before it and with
slightly different formatting. I have no idea who originally wrote that though.

Quotes of the week

Posted Mar 6, 2008 19:09 UTC (Thu) by iabervon (subscriber, #722) [Link]

Rusty also wrote it originally, on July 27th, 2007 (in commit 6c8dca5d).

Quotes of the week

Posted Mar 10, 2008 0:59 UTC (Mon) by jbailey (guest, #16890) [Link]

Poor device management is something that's confused me for a long time in Linux - almost
enough to want to scratch the itch myself, but not quite.

What I really want is what I had back in the old Novell Netware days - I had a logical volume
and a physical volume. At any time, I could:

* Add a physical volume as a mirror or additional space.
* Grow a logical volume across available space in the physical volume.

And this was all relatively straight forward. What surprises me is that with EVMS, LVM2,
mdadm, or any other fancy tools, I haven't found a way of making it so that I can configure a
system such that any system I start out with can suddenly become mirrored. There's always a
complaint about there not being enough room for the metadata, or some step that requires me to
have thought about what I might want to do in the future. And mirroring later on is my
biggest requirement. Suddenly I discover that a drive's worth of data is more precious than I
thought, so I want to add protection to it.

The NetWare solution was particularily cool in that it could do this with any set of block
devices it could see. This meant that setups like SFT3 were really natural, where disks were
mirrored across multiple machines.

Sadly, this is a space that AFAIK only Novell ever got super-right, in that it was always
there, you could always do it, and the commands to do it were simple enough that even book
CNEs could generally handle all the mirroring and remirroring tasks without worrying if they
were going to get the mirrors in the wrong direction.

Tks,
Jeff Bailey (Who wants volume management like we had it in 1993)

"How to not invent kernel interfaces" Arnd Bergmann, July 31, 2007 LinuxConf

Posted Mar 10, 2008 21:41 UTC (Mon) by dwheeler (guest, #1216) [Link]

Here's a link to the paper cited at the bottom: "How to not invent kernel interfaces" Arnd Bergmann, July 31, 2007 LinuxConf. It can be tricky to find with Google.

Quotes of the week

Posted Mar 15, 2008 7:58 UTC (Sat) by muwlgr (guest, #35359) [Link]

Indeed, Linux/Sistina LVM/LVM2 just omit LV mirroring (lvconvert -m), as well as a facility
for online filesystem&LV resizing (fsadm was there in LVM1, but not in LVM2). LV mirroring is
present in Veritas LVM (Solaris, HP-UX) after which Sistina's one was modelled, as well as in
IBM (AIX, OS/2) LVM. Under Linux, I still have to create lower-level mirror by mdadm, then LVM
PV over it.