Hello! So this is what I have read out on the web in the last week. As always, if I have missed something cool, then let me know in the comments.
Where are all the kittens?
Gentoo held a competition for new images to put on its screenshot page. Here are the results, pretty nice, although they all seem to be rather 'matrix'/'hackers' style, all a bit gloomy for me, I would have picked a wider range myself, i.e. mountains, landscapes, kittens, chicks, pre- renaissance art and so on.
Bash history
Brian Carper writes about how to keep bash history synchronised in real time.
This is pretty cool, however, while I do use my bash history for extra command completion, I think relying on something being in your command history is an inefficient way of working. Regular commands should be saved as shell scripts. Bash history also has privacy implications, so I prefer that it is wiped at regular intervals.
Pycon UK notes
Last weekend I was very busy during PyconUK, my first priority was to do my duties as an organiser and member of the crew, the second priority was to enjoy the conference as a delegate and go to as many talks as I could. Therefore, 'live blogging' the conference was not really an option for me. However I did learn a lot of new software and met lots of new people, so I'll probably write up my take on things later in one form or another.
However, many other people did blog at the time or soon after, including some fellow members of Birmingham LUG and Python West Midlands. Ciaran has a photo of me looking for all the spare plug sockets. Dave Goodwin has a nice summary of his personal highlights, and Simon Stanford has a more in depth, multi-part round up, starting here. A partial list of responses is available on the PyconUK Wiki.
Python presses onwards
A month ago, I talked about how I moved to Python 2.5 a little early. Well since then the package python-2.5.1-r2 is now in ~ (testing).
The world moves on and the alpha of Python 3000, Python 3.0a1, has been released, so Gentoo's Python team will soon have to start their compatibility audit of packages all over again!
Roeland Douma wanted to use braces in Python. I think braces suck as, in general, I believe in abstraction, i.e. computers should do as much of the work as possible, so within the spectrum of binary to English, programming languages should be as close to English as possible without losing clarity (i.e. the ability to say what you need to). Braces do not provide extra clarity so are redundant.
Anyway, it allowed me to roll out the old joke about how to use braces in Python:
#{
def hello():
print 'Hello World'
#}
hello()
Optimum bloat?
Bug has started to use XFCE as his window manager, he finds that it is the right trade off between features and system resources for him at the moment. Gnome being too heavy on system resources and Fluxbox too light on features.
With all graphical operating systems there is a trade off between how much system resources should be spent on maintaining the operating system itself versus how much should be left to use for applications. After all you turn on a computer to do tasks, rather than to have it on for the sake of being on.
Windows Vista, and Mac OS X are at one end of the spectrum. According to Microsoft, to use all the Vista features, you need 1GB of RAM. When buying a new computer for Vista independent Windows experts seem to recommend at least 2GB for a usable experience (4 GB for the 64bit version).
How much does Vista use when idle is a slightly difficult question to answer as Vista has a feature called 'SuperFetch' which uses up spare RAM guessing what you want to do next.
I have been surfing the Windows forums to find out the difference between OS usage and superfetch. A rather anecdotal approach, like a lazy reporter that uses 'a quick poll of taxi drivers', but there seems to be a consensus that somewhere over 500MB is being used by the core operating system, including background tasks but excluding superfetch.
I suppose the way to test this would be to take a number of Vista computers, floor them with a number of processes that take up 100% of RAM, then monitor how much of the RAM that is being used by your processes, the remaining RAM would be the operating system use.
I had one idea of how to pratically do this but it was flawed, see Bug in the comments.
Does anyone have an idea of how to measure this?
Why Darl is still grinning
I discussed SCO a lot recently, but this new week was supposed to be the start of the SCO vs Novell trial proper when the Judge would hear arguments about how much SCO really owes Novell, i.e. the money that SCO has taken in via licences but neglected to return to Novell. So we are talking about some proportion of $25 million dollars plus interest, perhaps also that again in legal fees.
It was starting to look like the proportion was going to be 100%, and IBM, Redhat are waiting for money in the queue behind Novell, possibly Autozone and DaimlerChrysler too. As I have explained many times before, from quite early on, say by the start of 2004, it was obvious, to some of us at least, that SCO never really had a chance, that IBM and Novell and had a "triple- lock" of defences. Since we have been able to predict for quite a while that SCO would lose and have to pay Novell mega money, these predicted losses have been effectively undisclosed balance-sheet liabilities, so SCO has been effectively operating while bankrupt for a while, a zombie.
So the trial was due to start on Monday, but SCO filed for 'chapter 11 bankruptcy' this last Friday. This will probably stay the case as we know it; however Novell might not leave it at that. Novell believes it is owed $25 million plus interest plus legal fees, this week Novell might have well been awarded the whole company in partial payment, however SCO has not listed this in its list of creditors in the bankruptcy papers.
Novell might have had enough with SCO and want to move on, but it should really push for itself to be first in line for the remaining scraps of cash, rather than not in line as the current SCO leadership have ordered the queue. This would mean Novell filing a motion for SCO to have chapter 7 bankruptcy instead and get a ruling from the SCO VS Novell Judge saying it is owed money.
Compared to Europe, America is very lenient to bankrupt companies. the two main forms of bankruptcy in America are called "Chapter 7" and "Chapter 11". Chapter 7 is a regular bankruptcy, a trustee is appointed who tries to get as much money as possible for the creditors - this normally means liquidation or sale. Chapter 11 is a bit different, this is aimed at giving companies breathing space to reorganise. The problem with this setup is that the executives that ran the company into the ground are still in place.
This is another way in which the American system has a tendency to shield executives from the consequences of their actions, the government intervenes in the market and the executives walk away to the next gig - scot-free - with their cash, exercised stock options and private pensions intact, while creditors, stockholders, employees and customers are left to pick up the pieces.
When Darl McBride took over as CEO of SCO, he could have gone with the previous plan, i.e. sell Caldera Linux as an upgrade plan to Novell's legacy Unix customers. This however must have seemed too much like hard work. Instead McBride concocted some crazy plot for SCO to become the gatekeeper for the whole multi-billion dollar open source software industry. Why create your own products and work hard for your own customers when you can leech off the work, products and customers of others? It was a gamble, that failed miserably, taking down the whole company in the process.
So maybe Darl McBride will be hired to burn down some other company. However, in 2003 alone, Darl McBride was paid $986,047 in cash plus stock options of 200,000 shares (source: marklogic). If he was paid a similar amount over 5 years, then he never has to work again if he doesn't want to, he can buy a ranch and relax. If he sold some shares when the stock price was high (e.g. when the stock was almost $20), then he can buy a yacht or two as well. So we won, but I doubt Darl McBride cares that much, he did quite fine thankyou out of the whole game.
If this is a bit depressing then let's listen to some blues, here is Andy Lockran (on bass - with the cheesy grin) and his mates playing 'Ain't no sunshine when she's gone'.
<p>Nicely written.</p>
<p>I think there is a flaw in your way of testing Vista's idle usage. I don't
think the infinite loop would use much RAM, but rather just the CPU.</p>
<p>P.S.
Teach me the art of track backs.</p>
<p>@bug</p>
<p>Crap you are right. What could we use? Some media encoding script?</p>
<p>On Linux we have <a class="reference external" href="http://www.berthels.co.uk/exmap/">exmap</a> that is good for using in memory testing, maybe
there is something for Windows that is similar?</p>
<p>A little while back I really looked into XFCE myself. I was trying to go
with the whole Desktop thing (I've been an avid Openbox user). I really
looked at the differences between Gnome and XFCE and I personally concluded
they were right on par with each other. I found that memory usage was about
the same and number of processes was similar. They also took a similar amount
of time to startup. I ended up back in my beloved Openbox, using Rox for my
filer. Just thought I'd give another impression on the Gnome vs. XFCE.</p>
<p>Hum, just checked out the exmap page, I've never heard of it before. It
might have been good to use in my XFCE trial. Very interesting indeed.</p>
<p>And on the memory subject. A media encoding script would be the same as an
infinite loop. Media encoding is typically CPU bound. You could have an
application which loops some number of times and in each iteration mallocs a
few megs of memory. It would also probably be best to insert some sleep in
each iteration so that it wasn't a tight loop. And finally, have it's main
argument be how MUCH RAM to grab that way you don't build up a ridiculous
swap file. That'd be my idea.</p>
<p>Definitely I don't keep anything critical in bash history. If I deleted mine
right now I wouldn't lose anything. I agree everything really important
should go in a bash script.</p>
<p>Bash history is good for 1) Relief of short-term memory, 2) Tracing your
steps to see what you did at some point in the recent past, 3) Being able to
play with commands at will until you realize that you keep doing something
over and over and THEN put it in a bash script, 4) Simply saving yourself a
bit of typing now and then (or a bit of copy/pasting between terminals).
Among other things.</p>
<p>Security is another concern, if your computer is multi-user, which mine
thankfully isn't.</p>