I’m rebuilding my main desktop system at the moment. It isn’t anything spectacularly new, in fact it is based on a Athlon XP 2800+, but it runs Linux (Ubuntu 9.10 currently) with ample performance for my day to day tasks. Windows XP runs on it adequately, when it actually works (I clearly use it too little because it sulks and ‘blue screens’ every few months, requiring a repair or reinstall).

Anyway, as part of this rebuild I decided that it made sense to get everything bang up to date, including the BIOS. The board is obsolete now, so the BIOS was a few years old and should be stable enough – wrong! The board is a Gigabyte GA-7S748-L (GA-7S748 with on board LAN), and was running the F5 revision. Updating was simple enough, just boot of my DOS USB key (yes, I have a USB key with DOS installed!) and run the utility. This brought things up to revision F9, which was the latest, and is where things started to go wrong.

The first thing I noticed was that my Windows 7 installation couldn’t see the HD, so I reached for a driver (Vista was the most up to date) to add during the install process. This didn’t work, and further investigation showed that the BIOS of the PATA PCI card the drives were connected to wasn’t active. It turned out that enabling USB legacy support disabled the card BIOS. Unfortunately, in order to configure the card BIOS you needed a keyboard, and since mine was a USB one, USB legacy support was required to use it – catch 22!! After trying a PCI SATA card instead I found that this was clearly a motherboard issue and not specific to one card or chipset.

So the next thing to do was to back off to the previous BIOS (F8 in this case) to see if the bug had been introduced in the latest BIOS upgrade. Unfortunately it seems that the F8 BIOS was completely bug ridden (so it is a shame it is still available for download). After installing it the computer wouldn’t boot at all, although it would allow use of the BIOS itself. This board has a facility to flash the BIOS from a BIOS based utility (Q-Flash), which sounds pretty handy in this case. Sadly it didn’t work, although thankfully, the failure didn’t damage the installed BIOS any further. After a bit of research with Google, I found that if you disable USB support completely you can boot to a floppy and re-flash the BIOS using a DOS based utility (will DOS ever die?!).

So, with this information I managed to get back to the F5 BIOS and all was working again. Out of curiosity I then tried the F6 BIOS and found that the USB legacy support clash with the PATA card BIOS was introduced there, so presumably exists in F7 too. Back to square one again then, with the original F5 BIOS and a working PC!

I note from recent news that Microsoft and Yahoo! have been given the go ahead for their search deal, which means that Microsoft Bing should be providing the search results for Yahoo! by the end of the year.

Interesting times, since that would appear to indicate that the average new user of Ubuntu will be using a Microsoft search engine, and one that may not be as good for searching for information on Linux and Open Source software – although it has to be said I’ve not experimented with this myself.

That said, let’s move on…

The other day my son came home and asked if he could use the computer. Nothing spectacularly unusual there, but shortly after logging on he came back to me asking if we had “the big blue W” on it somewhere. After a brief to and fro I ascertained that what he meant was Microsoft Word and that he wanted to show me something that he had done at school.

There’s a few points here. Firstly there is the comparison with when I was his age (he is seven). Back then I still had around 6 years until the ZX81 was launched and I got my first taste of computers. Now they are not only everywhere, but interconnected to the point where it didn’t even occur to him that the work he had been doing at school wouldn’t be accessible from home – bless him.

Secondly there’s the annoyance that from such an early age the ‘big blue W’ and Microsoft Word are synonymous with word processing. I keep trying to tell myself that he is only seven and they need to keep things simple, but it is still irritating. The end result of this was that I sat down and explained what a word processor was, and that even though I didn’t have the one he used at school I did have another one. I was encouraged by the fact that, when presented with OpenOffice he set straight to work trying to reproduce the same sort of document he had at school, and did so with no help at all – beyond a brief grumble that there wasn’t much clipart! (actually I did show him where that was).

The next question baffled me a bit, “do we have the internet?”. My response was to tell him that we did, and that he used it regularly! Then we headed down a familiar road, “no I mean the big blue e” So off we went down the explanation of what a browser was and that he was already quite comfortable using both Opera and Firefox, as was his younger brother (who is 3). In fact they are both rather too good at some of the games on the CBeebies, Nick Jr and BBC Dr Who websites! Still, they are also quite happy playing with Frozen Bubble or Tux Paint

To start with it is worth noting that I’ve only installed the Windows version, since I’m suffering using Microsoft on the desktop for a while. This is partly due to a couple of hardware failures, partly office reorganisations that are as yet unfinished, and partly forcing myself to keep my hand in – unfortunately I have to.

The first thing I noticed was that it was much faster and more responsive. Firefox 2 has, particularly in the past few weeks, been pausing for periods for no apparent reason. The second observation was that the memory footprint was much, much smaller, so this may well be linked. I tend to use a reasonable number of tabs when browsing and have been keeping an eye on the memory usage recently. I’ve been trimming things down to around 12 open tabs at a time and was amazed that Firefox was using between 200M and 260M, with no sign of reducing when tabs were closed. After installing Firefox 3 I checked and found that it was running at pretty much exactly half that, around 100M to 130M with the same 12 tabs as before the upgrade!

Part of this reduced footprint may be down to the few extensions that are now disabled due to incompatibility. With a major upgrade it is probably time to take a look through the extensions I’m using to see if any are no longer relevant. A few quick highlights of my extension usage are…

Bookmark Sync and Sort – sadly this is no longer compatible, and from a quick investigation it hasn’t been updated for a good while now, and the referenced home page has gone. Time to look for an alternative it seems. All this was doing was backing up my bookmarks to a WebDav folder on my intranet server, and at first glance there seems to be no obvious alternative. With a bit more digging I’ve found two, or perhaps just the one, options:

• Foxmarks – I ignored this first time round since it was linked to a service, but further investigation shows that you can use it with either WebDav or FTP, as well as install a server component. The original Foxmarks was based on the OSAF Cosmo server, but it looks as though this has been replaced by the Chandler project, so a little more investigation is required.
• Sitebar – another one I passed over when finding Bookmark Sync and Sort. Again it is linked to a service, but it seems that you can download the server component if you want to. Sadly, according the the Mozilla Addons site it doesn’t currently support Firefox 3, but it does have the advantage of being open source which Foxmarks doesn’t seem to be. Again, more investigation is required.

Firebug, IE Tab (remember this is on Windows!), It’s All Text!, Speed Dial and Web Developer are all available for Firefox 3, so that just leaves…

Tab Mix Plus – not having this one is a bit painful, although partly because I’m not sure what features I’m missing that were part of it and what may now be officially available that I haven’t found yet! That said, I’ve just noted that there is a newer version, just not listed on the Mozilla Addons site. I’ll have to give it a try and then work out what options I’m using that it provides – the session manager, the new tab button and undo closed tabs spring to mind.

There are a couple of other incompatible ones listed on my install. Google Photos Screensaver I think I had disabled anyway and RealPlayer Browser Plugin I hadn’t noticed before – although is probably responsible for blindingly easy downloads from YouTube etc.

That’s what leaps out at me, bar the slight interface changes. No doubt more differences will become obvious over time.

OK, with that out the way, I thought that I’d do a quick blog about the ease of installing .deb packages within Ubuntu (clearly I’m in a bit of a blogging mood today – something of a rarity!). I first stumbled across this a while ago when I decided to install the Opera browser alongside my usual Firefox install. I headed off to the Opera website to download the .deb file that I knew, from previous Debian installs, would be there. What I was expecting to do was download it, do a test install run with aptitude -s to check for any missing dependencies and then, after installing any missing items, install Opera – all via the trusty, familiar command line.

Things didn’t go as expected though. When I clicked on the .deb file that was provided I was given the option to ‘Open with’ the GDebi Package Installer. Even though I’ve been using Debian for some years now I’ve not come across this piece of software before. This is just another reminder, as if
one were needed, that there is always something new to learn however familiar you are with something, and however long you’ve been using it. Anyway, I decided to let GDebi do its stuff, and after downloading the .deb file it added a couple of dependency packages to the install list and proceeded (after a couple of confirmation prompts) to install Opera and the required extra packages – all nice and easy, and one more demonstration that, certainly to my mind, Windows has a long way to go to catch up with Linux in terms of ease of use

Today I was reminded of this easy install path for non standard software once again. Not to say that I had forgotten about it, but it isn’t often used. I was reminded, however, in a rather unexpected place. Once again I was off to download a piece of software to install, and yes, another proprietary piece, but some of us have to live and work in the real world where, unpalatable though it may be to some, we have to use proprietary software – and to be honest, I have to concede that there is likely to have to be a place for it in some cases, but that is getting off topic. I was off to the Adobe website to download the latest copy of the Adobe Reader (or Acrobat as it was once called). What I was expecting was to download the tar.gz file, extract it and then use the installer. I’ve done this before, and have tested pre-release versions in the past. One of my feedback comments was that it would be nice to have a .deb package version as well as the .rpm one. Well I can’t say how much impact that feedback had, but I can report that there is now a .deb package for the Adobe Reader. I suspect that the rise and rise of Ubuntu probably has at least something to do with this. Whatever the reason, I was presented with a .deb file, which then installed with ease using the GDebi Package Installer. Once again had a pain free installation on Linux – no battling with dependencies (as per the ‘olden days’ of Red Hat), no bits of my system being over written with different versions of files that break existing applications (as can happen with Windows, mfc42.dll anyone?), just a few clicks and a pain free install.

After using a pretty standard Ubuntu install with Gnome for a while I decided to switch back to my preferred XFCE desktop. This was fairly straightforward, although the xubuntu-desktop package pulled in a few things I didn’t want, so I used it to populate the new install list (in Synaptic) and then trimmed the packages I didn’t want (mainly the office applications since I prefer to stick with OpenOffice). One thing I noted after making the switch back was that there was no Network Manager applet running, and a quick check of the package details seemed to indicate that it was part of Gnome, so I decided to look elsewhere for an equivalent. I experimented briefly with Wicd, but whilst doing that I found out that, although Network Manager is part of Gnome it is pretty free of Gnome dependencies and is largely independent, so I decided to switch back. Having done that I needed to look for a way of getting the applet running. This seemed simple enough, I just needed to add an entry into the Autostarted applications list as accessed from Applications -> Settings. Having done this all seemed well until, after a few reboots I noticed that I had multiple copies of the nm-applet in the panel – oops!

After a bit of digging around my file system in directories like /etc and ~/.config with the grep command I finally came up with some likely files, mainly /etc/xdg/autostart/nm-applet.desktop. Then with a bit of help from Google I came up with a couple of good references, mainly a bug report for Ubuntu [1], a reference on the Ubuntu forums [2] and another discussion linked from there [3]. With the aid of those and a bit of experimentation I came up with the following fix:

1. First off I got rid of all the nm-applet instances running with pkill nm-applet
2. Next I got rid of my hand installed autostart entry for the nm-applet, so to do this I deleted the file ~/.config/autostart/nm-applet.desktop
3. Then I made an edit to the system wide autostart entry for nm-applet, for this I need to change a line in the /etc/xdg/autostart/nm-applet.desktop
   file from OnlyShowIn=GNOME; to OnlyShowIn=GNOME;XFCE;. This lets the nm-applet configuration show in the Autostarted applications utility when you’re running XFCE.
4. Finally I needed to clear out any information relating to nm-applet in the cached sessions. This was a case of editing two files in the ~/.cache/sessions/ directory.
• These were xfce4-session-dusky:0 and xfce4-session-dusky:0.bak dusky is the name of my machine, so your files will have your machine name instead).
• These files have a number of lines, the majority of which are grouped with the prefix Client0, Client1, etc., so I looked for the groups with an entry for CloneCommand of nm-applet and removed the groups.
• I then had to renumber the Client groups so they were consecutive again starting at 0 and amend the Count entry towards the end of the file to match the number of groups I had (so one more than the last number used, remembering that you started at 0!).
5. After the editing I could perform a reboot to confirm that all was well. I actually did a few of these, firstly with the Save session for future logins disabled to ensure that any cache information was completely clear. Next was another with the session disabled, just to be sure, and then one with
   it enabled again.

All now appears fine with just a single nm-applet on my panel.

[1] https://bugs.launchpad.net/ubuntu/+source/network-manager-applet/+bug/95064
[2] http://ubuntuforums.org/archive/index.php/t-191708.html
[3] http://www.nabble.com/Multiple-Instances-of-nm-applet-t4114321.html

As I suspected, the Linksys WMP55AG worked a treat and was automatically detected and, after a bit of a pause while I cursed it not working, my two local networks appeared in the network manager applet and off I went. It correctly identified that I was using WPA and allowed my to put my TKIP passphrase in and off it went. A quick install of Thunderbird and it should be ready to start pursuading my wife to give it a go. It will be interesting to give this installation a real workout as the Windows XP install on the same machine runs like a dead Norwegian blue parrot!

I guess I should give a bit more detail, but for now, since I’m tired, I’ll call it a night. I will add that the card uses the MadWiFi driver which is manged by the restricted-manager utilty – I’ll have to learn more about that now so I can get the nVidia driver working properly.

So, where am I with my selection of cards? Somewhere about here:

• 3Com 3CRWE154G72 version 1.0 PCMCIA
  This did look to be my best bet, it uses the prism54 driver which on first glance looks to be supported by wpa_supplicant, sadly however, if you read closer you realise that the driver is only supported by hostapd and not wpa_supplicant – which to me seems odd, but there you go – so I assume this means it could be used as an access point, but not as a client! – Ubuntu identifies this OK and uses the prism54pci driver.
• Linksys WPC54G ver. 4 PCMCIA
  This has got off to a very bad start, even though it uses the rt2x00 chipset which should work, still no good on WPA though sadly – Ubuntu doesn’t seem to be able to identify this at all, it is listed as an ethernet controller: unknown device 07fe:2220.
• 3Com 3CRWE254G72 USB
  No idea on this one, and I’ve not yet identified the chipset – Ubuntu has no idea that it is a network adapter or what driver to use with it.
• D-Link DWL-G122 Rev. B1 USB
  Another rt2x00 driver based card, seemed to have a good start, but again no WPA support – Ubuntu seems to load the rt2570 driver for this one.

Later I hope to move onto the PCI cards in a desktop machine, so will be looking into:

• Linksys WMP55AG PCI
  This looks as though it may be the best of the lot with an Atheros chipset supported by the Mad WiFi driver!
• MSI PC60G PCI
  Not a clue on this one yet, still in its shrink wrap.
• USR 5416 PCI
  Looks to be based on the ACX 111 chipset, beyond that I’ve not investigated.

I’m not entirely sure whether Ubuntu should support WPA out of the box in 7.04. All the documentation on the wiki refers to earlier versions, but I suspect that all the pieces are in place but need a bit of manual hacking in configuration files. When (there’s hope for you) I get it working I’ll have to document it. It looks like ndiswrapper almost all round though sadly.

Speaking of documenting things, I’ll make a quick note here that I need to do just that on a couple of things:

• virtual users and virtual domains with Exim 4 and Dovecot
• Synchronizing and backing up Firefox bookmarks with Webdav

• first off I downloaded the file: http://www.mplayerhq.hu/MPlayer/releases/codecs/essential-20050412.tar.bz2 from the MPlayer site
• next I extracted it and renamed the directory it created to codec
• this new directory was then moved to /usr/lib/
• finally a symlink was created so it also showed as /usr/lib/win32/

Not entirely sure whether working with it symlinked is necessary, or whether simply creating it as win32 would have done. It works OK though, so I’m not really complaining!

So why haven’t I linked the download? Well I’ve just checked it and there’s been an update, so this looks to be the latest version:

http://www.mplayerhq.hu/MPlayer/releases/codecs/essential-20060501.tar.bz2

So I’m downloading that for an update!

First off the Device sections. I had an original one for my NVidia card along these lines:

Section "Device"
        Identifier      "NVidia GeForce MX 440"
        Driver          "vesa"
#       Driver          "nv"
#       Driver          "nvidia"
        Option          "UseFBDev"              "true"
        BusID           "PCI:1:0:0"
EndSection

Note that the two commented out lines are due to switches back and forth with the NVidia drivers. So far the vesa one is the only on that has consistently worked and for my usage is doing me fine for now!

I’ve added a second Device section for the secondary card:

Section "Device"
        Identifier      "Matrox Millenium II"
        Driver          "mga"
        BusID           "PCI:0:10:0"
EndSection

One change since my first problematic configuration was to switch to the mga driver, but it terms of problems this change didn’t appear to impact anything. Note that the BusID lines are pulled from information in lspci:

0000:00:00.0 Host bridge: Silicon Integrated Systems [SiS] 746 Host (rev 10)
0000:00:01.0 PCI bridge: Silicon Integrated Systems [SiS] SG86C202
0000:00:02.0 ISA bridge: Silicon Integrated Systems [SiS] SiS85C503/5513 (LPC Bri
dge) (rev 25)
0000:00:02.5 IDE interface: Silicon Integrated Systems [SiS] 5513 [IDE]
0000:00:02.7 Multimedia audio controller: Silicon Integrated Systems [SiS] Sound
Controller (rev a0)
0000:00:03.0 USB Controller: Silicon Integrated Systems [SiS] USB 1.0 Controller
(rev 0f)
0000:00:03.1 USB Controller: Silicon Integrated Systems [SiS] USB 1.0 Controller
(rev 0f)
0000:00:03.3 USB Controller: Silicon Integrated Systems [SiS] USB 2.0 Controller
0000:00:04.0 Ethernet controller: Silicon Integrated Systems [SiS] SiS900 PCI Fas
t Ethernet (rev 90)
0000:00:0a.0 VGA compatible controller: Matrox Graphics, Inc. MGA 2164W [Millenni
um II]
0000:00:0c.0 SCSI storage controller: Adaptec AHA-2940/2940W / AIC-7871(rev 03)
0000:01:00.0 VGA compatible controller: nVidia Corporation NV17 [GeForce4 MX 440]
 (rev a3)

Looking at the last entry in that list you can get the 1:0:0 for the AGP card (not entirely sure why you use PCI in the line, likely just a legacy naming issue for the config file). The PCI card comes in the third line from the end and gives 0:10:0 since 0a is in Hex and translates to 10 in decimal. These BusID lines are required with more than one graphics card, but not with a single card.

Right, on to the Monitor sections now. I have two of these that are identical bar the Identifier line which has a #0 and #1 for each:

Section "Monitor"
        Identifier "Iiyama Vision Master Pro 400 #0"
        VendorName "Iiyama"
        ModelName "Vision Master Pro 400"
        HorizSync 27-96
        VertRefresh 50-160
        Option "DPMS"
EndSection

Since the Iiyama screens didn’t respond to the get-edid command with good enough information to pipe through to parse-edid I took a quick look at the auto generated section for the previous monitor and matched the only lines not commented out – obviously with the necessary specification changes!

I’m not entirely sure whether you need two sections if the monitors are identical, but as a problem solving process I tried it and haven’t tried changing back yet!

Right, twin Screen sections now. I’ve trimmed these drastically to ensure both screens are working at the same resolution and colour depth (which is apparently required). From the checks I did and the configuration they would have been identically anyway, but again this was a process of elimination:

Section "Screen"
        Identifier      "Primary Screen"
        Device          "NVidia GeForce MX 440"
        Monitor         "Iiyama Vision Master Pro 400 #0"
        DefaultDepth    16
        SubSection "Display"
                Depth           16
                Modes           "1280x1024"
        EndSubSection
EndSection

The second one is pretty much the same bar the top three lines which are:

        Identifier      "Secondary Screen"
        Device          "Matrox Millenium II"
        Monitor         "Iiyama Vision Master Pro 400 #1"

Finally there’s the ServerLayout section which reads much like this:

Section "ServerLayout"
        Identifier      "Dual Screen"
        Screen          "Primary Screen"
        Screen          "Secondary Screen" LeftOf "Primary Screen"
        Option          "Xinerama"
        InputDevice     "Standard Keyboard"
        InputDevice     "MS Optical USB Mouse"
EndSection

The key change I’ve made here is in the Screen lines. Initially they were typed based on an article I was reading on the subject on the Linux.com website:

http://www.linux.com/article.pl?sid=03/10/05/025207&tid=13&tid=39&tid=23&tid=99

This had the information presented in this form:

        Screen          "Secondary Screen" LeftOf "Primary Screen"
        Screen          "Primary Screen"

However switching those two lines to make them read more logically gives:

        Screen          "Primary Screen"
        Screen          "Secondary Screen" LeftOf "Primary Screen"

So there we go, it works I still have to decide on a resolution to run at (the Matrox card is sadly lacking in performance.