DevHeads.net

Here's a thought I had about Grub 2 based on recent discussions on this
list.

I really love Grub2, and think it's been a great improvement in just
about every way. But the automatically generated grub.cfg is just not
always easy to cope with for people with more complex systems (ie, maybe
someone has multiple versions of Windows and Linux...)

However, if you rely on 40_custom for your grub.cfg, you then have to
update that file yourself to benefit from kernel updates.

My idea is that is that a file like 45_ubuntu_custom be created in
/etc/grub.d but be made non-executable.

As I understand it, (I might have some facts wrong)

They are trying to no longer rely on BIOS drive ordering (because they
change all the fricken time) and blocklists (because they can break if
someone does something as simple as defrag a filesystem, and all of a
sudden grub stage 1.5 files are no longer on the same blocks.)

So, in theory, grub should now embed enough executable code that it can
read the filesystem that rest of it is located on, and it can search all
hard drives the bios is aware of until it finds those files, at which
point, grub.cfg (the replacement for menu.lst file

I couldn't help but notice recently, much to my cost, that Intel is
removing Virtualization extensions on all their current socket 775
CPU's.. (how are the i5 and i7 faring?). The only reason I can think
for this move is to prevent the low end cpus from competing with the
workstation/server market, since these cpu's are now more than fast
enough for many workloads to be simply io bound.

I was trying to install build dep packages for gxine (since the one
included with Karmic is in really bad shape.) and stumbled on this
package. libmozjs-dev requires un-install firefox (and 18 other firefox
related packages,) which makes no sense to me. I will file a bug
report, but first wanted to ask the question here in case there is, in
fact, a reason for this package to be like this.

Why would you even bother translating hex to decimal in an IP Address?
the numbers have no meaning unless you translate them to binary
(because the subnet division doest have to be an an even octet, even
though most people always do so.) and co-incidentally, it's *much*
easier to translate hex to binary 'on the fly' (that's why we use HEX
in the first place.) Having to convert Ip addresses to dec. has always
been a ridiculously unecessary and unhelpful flaw of the convention.

Now that X no longer uses an xorg.conf file by default, how is the
default Keyboard map determined?

As I discussed earlier in my rant thread, PolicyKit has been replaced
with PolicyKit-1, but there is now no gui to configure any of the policies.

It was easy enough to find the /usr/share/polikit-1 folder and I can
modify the entries there to change the deault policies to my heart's
content, but I can't figure out how grant access to a specific user.

For a real world example, suppose I want to give my primary user
permission to use sound devices, whether or not I'm currently the active
console user.

And no, I'm not complaining about some bug that wrecked my install or
upgrade. At least that part seems to be really smooth.

But the new gnome is a wasteland. Many components have been updated
that are not close to being ready for prime time. Some examples:

GDM - lots of threads already about how hard it is to control the face
browser. What about enabling xdmcp, or disabling suspend/hybernate on
systems where does might cause trouble? (without remove the shutdown
buttons entirely, though that might be the best option).

How can I disable apparmor in Karmic?

I tried changing the rcS.d link from S37apparmor to K37apparmor (as I
have in previous versions of ubuntu,) but that didn't kill it as advertised.

I then tried to remove the symlink entirely, but that didn't work
either. (presumably, it's started right from initrd?)

What is the recommended best practice approach to disable services
controlled by upstart (in the /etc/init folder rather than /etc/init.d)

Is there a website where we can post Ubuntu ideas / feature requests? I
remember some kind of "brainstorm" site, but I think that was a one time
idea gathering.

I think it's long past time that network manager get some kind of plugin
to enable/disable ufw (ie, Microsoft's trusted vs public network) for
different network connections. That's one idea they at least got right.

Here's a phenomenon I've observed in all Ubuntu releases starting with
Gutsy, and I've never been able to figure out an explanation. Can
somone explain to me why transferring a file over sftp protocal with
Nautilus only works half as fast as using command line sftp client or
filezilla?

I've noticed this both when transferring files from a distant server
over internet or a server the next room over on Fast Ethernet Lan. If I
use sftp over command line or Filezilla, the network bandwidth used and
file transfer speeds are almost double.

I would like to file a bug/wish for the Gnome Main Menu. (The one
that's a single icon, as opposed to the Ubuntu default menu bar)

Can someone tell me which package this would belong too. Gnome-panel or
gnome applets?

Is there yet an open source, or at least, freely available Midi
Synthesis for Ubuntu? I still have the so called Gus Patches that I
downloaded years ago for Timidity (though I haven't tried installing
them in a modern distro since Slack 7), but surely there must be a
better way of doing this by now?

You really need to play closer attention to NTFS.. it's not nearly as
intelligent as you think. NTFS is the only file system I know of that
will fragment a new file, into hundreds of fragments, on a HD that is
almost all contiguous free space.. it's just flabbergasting.(tm)

I'm trying, unsuccessfully,. to find a utility that will display the
memory allocated, not including cache and buffers.

The closes I could come is with the free command. I can add the -/+
buffers/cache and Swap used column, which should be fairly close.
However, those numbers will contain 'clean' pages that exist both in
physical ram as well as used swap space.

Actually, $8400, (only the Pro version counts as Exchange replacement
with Outlook clients)

Still, much less pocket change than the 12k, so you're right, they've
adjusted the pricing.

875 * 5 years (I usually deploy a server for longer) = 4375. You need
to pay annually for the commercial version of Zimbra that includes the
Outlook connector. And you really don't want to the same comparison
for, say, 30 users :)

Microsoft Support? you'd have to be a sucker to buy into that :) But I
agree that cost of server + server Cal's is the devil in the details..
that's why I said earlier that Exchange is not expensive *if* you
already have server and Active Directory for your users.

Yes, that's exactly what they are for... More precisely, they are
symlinks to the scripts in init.d

That's how it's supposed to work. The scripts you find in ini.d are
controlled by the entries in rc?.d, so different scripts are fired at
different run levels. In this case, runlevel 2 should be the only one
you need to worry about unless you've customized things otherwise.

I used to use a perl script called conntrack-viewer that generated very
nice reports showing me all the NAT connections.. (Ie, I could see which
computers in my network were making what connections to the Internet).

Alas, this was designed for kernel 2.4, and doesn't seem to work with
Ubuntu Hardy 2.6 kernels.

Can anyone suggest an alternative program to do the same thing? thanks

Umm, I meant application/octet-stream. and it seems it doesn't matter
what the file name extension is. I can call it test.exe and IE still
opens it in the browser if it's html.

IE must be sampling the contents of the file to determine it's type if
it does not have an application associated with a given content type.
But in that case, the PHTML file that started all this should have
worked.... so is this behaviour only for application/octet-stream?