kontrawize

I beg your pardon, I never promised you a Rosegarden, part 3

Well, here's the funny thing. I wrote previously about how I had been having problems getting the Rosegarden sequencer and music editor to run under SUSE Linux 10. I haven't been able to find a fix for the problem as yet, but what I did discover is that it is mainly one of the supplied samples (unfortunately, the one I always use as my first test) which displays the problem. Many other pieces are not noticeably affected. So, I'm back using Rosegarden again, the latest version from CVS compiled on my laptop under SUSE 10.

I still highly recommend Rosegarden. Certainly it's given me a great way to write music using regular musical notation, which is the way I learned to do it at school. Note I said a great way to write music, rather than a way to write great music, but it's just something I do for my own interest and edification.

By the way, for editing digitised music samples (rather than MIDI sequences), I've started using Audacity, and it's pretty neat. It's also available for Windows and Mac, so it's probably the best (read only) multi-platform waveform editor available.

Anyway, I have things working again. It just remains to wish you all a Happy New Year! Have a great one.

Fluent Interfaces, part 2

John Cowan commented that my critique of the impact of fluent interfaces on software debugging was based on an assumption that debugging would need to be done in a post-mortem fashion, rather than in an interactive fashion as the code was written. He is, indeed, completely right.

I can see how fluent interfaces would have been fine for Smalltalk, because nobody ever (I suspect) wrote an application server in Smalltalk. I've had to do a lot of work recently with Java application servers, and debugging is a huge issue. When something goes wrong in a live production system, you don't necessarily have the opportunity to run the entire application on your desktop PC in a debugger and (if you are very lucky) recreate the problem. What you have are the audit logs that your application has written, possibly including a stack trace of one or more exceptions.

The difficulty is that the kinds of systems that impose such restrictions are also typically high value, high risk systems, in the sense that software failures can be very expensive to rectify. Failures may have financial implications for clients of a company, and the company itself could be subject to fines or lawsuits. It does give you a completely different perspective compared to what is appropriate is you are developing a small desktop application, like an MP3 player, where a failure would be a nuisance, but probably not have financial consequences to the actual end users.

So yes, please see my comments in that light; big systems impose big restrictions on how you can write code in order to be able to properly maintain those systems once they get into production.

[ Technorati tags: software architecture, Java ]

Fluent Interfaces

Martin Fowler has written an interesting piece on Fluent Interfaces, which are APIs that are designed to make the code easier to read. Essentially, they do this by removing the need for lots of intermediate local variable declarations and assignments.

Pity, though, that he didn't discuss debugging. The usual reason why people ban this style of coding in because every fluent line of code depends on multiple method calls each returning a valid object. If any of the method calls returns a null value, the line of code fails. The problem here is that if you look at your debugging trace, you (hopefully) will find out which line of code was responsible, but you may not be able to trace which method was responsible. Even if you know which method was responsible, it can be difficult (depending on your debugger) to quickly locate what each of the return values are, and which one is returning the null. It can also be hard to evaluate an expression that uses any of the intermediate objects, since they don't have names that can be used in your expressions.

I'm not against fluent programming, but it's important to realise that it's a style that needs good tool support to make it manageable, especially when your focus moves from writing the code to debugging and maintaining it.

[ Technorati tags: software architecture, Java ]

I beg your pardon, I never promised you a Rosegarden, part 2

I wrote that Rosegarden doesn't run properly in a SUSE Linux 10 VMware session. Well, it turns out to be rather different to what I thought.

After that post, I installed Rosegarden directly in my host SUSE 10 installation, i.e. not in VMware. Guess what? I had exactly the same problem. So VMware wasn't the issue; in fact, VMware did a perfect job of replicating the behaviour on my physical laptop. So what was the problem?

I'm still in the process of tracking it down with the help of the Rosegarden developers on their mailing list, but the thing we are looking at is the system timer resolution. It used to be that Linux systems had a timer resolution of 10msec (milliseconds). They could not time anything more accurately than that. This is not good enough for MIDI, since people can hear that size of delay in a piece of music. Recent Linux versions (and I assume this included SUSE Linux Pro 9.3) reduced the resolution to 1msec, which is great for audio applications. However, there were apparently some issues for some applications with the ten-fold increase in timer-related CPU activity.

It seems that in SUSE Linux 10, the system timer resolution is now 4msec. While this is better than 10, it's a lot worse than 1msec. This may be the reason for the problem with Rosegarden, but it's far from certain as yet. The jury is still out; I'll let you know when and if there is a solution. I certainly want and need a solution.

If you are American, the US is spying on you

Just read a story in The Register that the US government has been caught using the ECHELON system to spy on its own citizens. For those of you who haven't heard of it, ECHELON is a super secret (yeah, right) computerised spying sytem that intercepts radio transmissions and telephone calls, automatically interprets the speech, and searches for words and/or phrases that are of interest. It has a huge capacity, so it isn't just intercepting random transmissions, it is intended to intercept all transmissions. It is run by English-speaking nations, so the UK, Canada, and Australia are also involved (not sure about others).

While the other nations help in collecting the intelligence, it seems that the US holds on to the actual results fairly tightly, so that the other nations don't get the same access that the US has. I remember media reports some years back that the US had been caught using ECHELON to spy on its partners Canada and Australia as part of its industrial espionage operations. Allegedly, information from ECHELON on what Canadian and Australian companies were going to bid in commerical negotiations was passed to US companies so that the US companies could undercut them.

You can understand how galling this is to Australians (like myself), since Australia considers itself to be one of the US's closest allies, and has followed it dutifully into every war since World War 2. So to have the US use this system, one that that Australia supplies information for, to conduct industrial espionage against Australian companies is very sad indeed.

Anyway, Americans get to share in this too now. I'm not sure if that makes me feel better or not, in retrospect. It will certainly make it easier to stage new-age re-runs of Watergate. Oh well, at least the US doesn't consider itself an international role model, or the moral leader of the free world. Oh, wait ...

I beg your pardon, I never promised you a Rosegarden

I've mentioned previously how impressed I am by Rosegarden as a tool for composing music. I was thrilled to find it included with SUSE Linux Pro 9.3, so I was somewhat gutted when I realised that it wasn't included with SUSE Linux 10. I looked around the Web, but no-one had created a suitable Rosegarden RPM installer for SUSE 10.

Next I downloaded the source RPM from SourceForge, but it wouldn't compile without errors. I'm guessing that as SUSE Linux goes from one version to the next, packages that don't compile are simply dropped from the distribution. That makes sense for them, but it's a potential pitfall if you upgrade your SUSE Linux and then find that you are missing an application that is important to you.

I found a recommendation to download the latest Rosegarden sources directly from the CVS repository at SourceForge. I did that, but found I also had to download a number of supporting packages (and some newer versions than are supplied with SUSE 10). I also had to install scons, which is a new-age Python replacement for make (and hence a bit of a competitor to Apache Ant and Apache Maven). However, with all of that, I was able to compile Rosegarden and get it running.

One more piece of bad news, at least for me - when I fired up Rosegarden in a SUSE Linux VMware session, Rosegarden complained that precision timing was not available. That confirms my previous (less than surprising) conclusion that VMware isn't ideal for running multimedia programs (at least for Linux sessions, it's worked OK for me running Windows XP virtual sessions). So I am going to have to build Rosegarden all over again in my host copy of SUSE Linux 10, rather than in my virtual session. To be fair, multimedia Linux in a virtual hardware environment is pretty bleeding edge stuff, so I'm not going to complain too much that it doesn't work. Still, give it a year or two, I'll probably give it another try. Might give Xen a try at that stage too, see if it can support the full spectrum of the things that I need and like to do.

My Pick of Web 2.0 - GarageBand

There's a lot of hype and hysteria around the Web 2.0 idea at the moment. Not that it is clear what it is, but the best definitions I've read suggest that it is all about people being involved and making content for the Web, where Web 1.0 was more about people just reading anything they could find using Google et al.

So I found a site today that I guess qualifies for Web 2.0 - GarageBand. It's a place where new bands can upload their music for everyone to hear, and listeners and rate and review their efforts. I think it's way cool that I can listen to so much new music, and I like the idea that bands can make a fairly direct connection to their potential audience.

At the moment I'm listening to some songs from Intervision 5. Smooth jazz/rock, with overtones of Jamiroquai. Well worth a listen.

SUSE 10 and Oracle 10g XE (Express Edition)

Many of you will have read that Oracle has released a free Express Edition (XE) of its 10g database. It's fully featured, but with some capacity limits built in to restrict it to smaller datasets and/or development use. For me, as someone whose clients often use Oracle databases, it's great to have access to this kind of implementation for software testing (I don't do enough relational database work to justify buying a full copy of Oracle just for my testing purposes).

Bad news for me is that the RPM installer for Oracle XE doesn't work properly with SUSE Linux 10. It's good that Oracle is embracing RPM installers now (I believe this is the first time for an Oracle database), but it seems that it's been done in a way that works for Red Hat Linux but not for SUSE (which is the other big one). It seems to be a permissions problem, so it seems it may be fixable (but no time to try it today). If you download Oracle XE and join the special forum, you can check out this thread on the issue. Unfortunately, you have to download XE and register to get access to this forum.

The Virtual Laptop, part 5: tick tock

Well, in spite of what I wrote in part 3 and part 4, my attempt to get the SUSE 10 clock to run at the right speed in a virtual VMware session hasn't worked. Setting the VMware host.cpuKhz setting to 1700000 (1.7GHz) didn't work as I initially thought. I was fooled because the behaviour seems to be different depending on whether I resume a suspended virtual session, or whether I boot it up from scratch. If I boot up, the time difference between the host system and the virtual session is definitely worse.

As an experiment, I tried setting host.cpuKhz setting to 1800000 (1.8GHz), thinking that it should slow the clock in the virtual session. However, there was no obvious slowing (or speeding up), so I'm back to square one. For now I'm basically using John Cowan's solution, and switching on the SUSE NTP client just long enough get the time right. I don't want to leave NTP switched on, as it would try to sync the time every time my laptop boots, and I'm generally not network connected when I boot it up, so I would just be waiting for a network timeout every time.

Tick tock, tick tock.

Laptop battery life

Stan Dyck asked why completely discharging a laptop battery would help keep it in peak condition. As many of you would know, rechargable batteries tend to suffer from a memory effect, where they tend to charge up only to the highest level they were last charged up to, and discharge only to the lowest level they were last discharged to. The trick for battery manufacturers has been to engineer batteries to avoid this effect. Batteries are much better now than they were, but it is still the case that if you tend to only partially discharge your laptop battery before recharging it, you will probably find that the maximum battery life will decline, perhaps markedly.

The best example I saw of this recently was with someone who has an IBM Thinkpad. The battery life had dropped to less than an hour. IBM provide a neat utility that gives you a constant view of the state of the battery. Using that as a guide, the owner was able to discharge and recharge the battery more fully each time until it got back to a two hour battery life. Maybe this problem will go away when we get laptops with fuel cells, but until then, it's still a good habit to discharge your battery as much as possible (or practical) before recharging it. For my part, I usually try to wait till mine is down to 10-15% of its peak charge before recharging it.