Then again there is the huge job of learning all the common libraries and services languages have. Consider this: I’ve used Java on and off for various little programs at the university over the years, but it has been so occasional that I still need to refer to documentation for a lot basic stuff like accessing files etc. It is kinda tough when you don’t have a very good memory for stuff like that. Similarly for the last year and a half I’ve needed Perl steadily, but occasionally on my job: I really need the documentation to check pretty basic functions.

The point I’m rambling to is that routine is in my opinion the biggest factor in mastering a language. If you don’t use it, you lose it. Basics may stay, but the finer stuff goes away.

My current part time employment is coming to an end and yet again I’ve been wondering about applying to programming jobs. The thing is, I’d really like to finish my thesis first. The problem is that while I finish my thesis I lose most of what little routine I have for programming. That makes it tad more difficult to ace interviews in the winter. Luckily I’ll propably have couple of odd jobs at the Library during the summer, but those are more likely to be configuring software rather than developing it.

It’s pretty horrible user experience that shutting down the console at wrong moment you can destroy hours of work accidentally. The correct way would be to have at least one backup autosave slot for the same gaming continuum to make sure that shutdown doesn’t completely wipe out player’s progress in the game. It isn’t end of the world the player is thrown back to the checkpoint before last.

Compare this to for example Fallout 3, which autosaves every time you change between game areas. It always does a fresh save. Even if the last save gets corrupted you can load the previous one. Of course Fallout 3 also has manual save option, which is even better for security(and user control).

In general autosave is a very nice idea from user experience point of view even if it takes away some control from the user. From game developer perspective it is pretty good option to make your game to be played in a certain way. However, it does mean you need to be very careful that in no checkpoint the player can be in so dire straits that continuing would be extremely difficult.

Of course this is parallel to all the stuff I’ve been doing for my thesis. In usability testing, card sorting and heuristic evaluation you do find usability problems with just one test subject, but with more patterns start to emerge and the data you get is more meaningful. Similarly I naturally believe that the research portion of my thesis would be better if I did as a part of a team or with a partner.

I definitely prefer working in teams even though I’ve worked mostly independently. It’s a general win-win, in a working team you generally improve every member in the two most important aspects: skill and motivation. Can’t for the life of me remember whether it was Rands or Scott Berkun who wrote about the skill and will as the aspects manager must massage.

# groupadd mguest
# groupadd media
# useradd -g mguest mguest
 
# zfs set aclinherit=passthrough tank/media
# zfs set aclmode=passthrough tank/media
 
# /usr/bin/chmod -R  A=\
> owner@:rwxpdDaARWcCos:fd-----:allow,\
> group:media:rwxpdDaARWc--s:fd-----:allow,\
> group:mguest:rxaARWcs:fd-----:allow,\
> everyone@:rwxpdDaARWcCos:fd-----:deny \
> /tank/media/

So basically every read only device uses the mguest user account and anyone who needs to edit the stuff belongs to media group. For some reason

group:mguest:read_set:fd:allow

didn’t work for read access.

One catch is to remember to use /usr/bin/chmod and not /usr/gnu/bin/chmod as the latter doesn’t support ACL.

The cloning is as simple as:

1. download Disk2vhd and extract it to some folder.
2. Start the software and choose which partition you want to clone.
3. Wait for the cloning to finish, in my case this took about 50 minutes
4. Copy the resulting .vhd file to your virtual machine host.
5. Start your virtual machine software and make a new VM.
6. Mount the disk on the VM you just made and boot.
7. If everything went right you should now have a clone of your Windows system running in a VM

I just tried this and just like that I have my old Windows 2008 running in a VirtualBox virtual machine on top of OpenSolaris.

If you are migrating a Windows installation to VirtualBox you might want to check out the how-to on on the VirtualBox website.

At first I intented to buy only 2 more hard drives, adding maybe 2-3TB. Then I got this idea wouldn’t it be fun to do away with all the separate disks and handle all the storage as one pool. This lead me to ZFS and this wonderful write-up about it.

Hardware choices

While I haven’t bought anything except sil3124-based sata raid card and couple of Dell raid cards($27 for the 2, surprised it they work), but this is how the escalation has gone so far:

1. Buy 4 1,5TB drives to use in RAIDZ configuration added to the existing server
2. Upgrade desktop to Intel Core i5-750 and use the E6600/4GB for the fileserver
3. Get Xeon/server mobo/4GB ECC RAM and a huge case for the fileserver

I’m basing the current choice on the positive idea that if I don’t need the fileserver to be that beefy I can use it as a desktop. It’ll be interesting to see whether Intel Lynnfield Xeon with power management uses less power than the AMD 4850e. Tested the current server with OpenSolaris livecd and it idles at 100W, compared to idling at 50W when using WS2008.

Hard disk choices

There is no perfect choice for home use if you intend to use any sort of RAID. This is because of TLER. Good discussion of it found at OpenSolaris forums.

1,5TB is currently the sweet spot price-wise as you can get Samsung or Western Digital drive for roughly €85-90 from various German online shops. However these drives aren’t rated for RAID use and in error situations can become unresponsive which in turn can result in them getting dropped from RAID ARRAY. Pretty much all the drives rated for RAID operation cost roughly 30-50% more. That is a huge premium in my case because I would pay for the extra performance that I wouldn’t need.

This basically means that I’m propably going to try my luck with the consumer class drives as the cost is an issue. 6 1,5TB drives cost roughly €520 compared to 8 1TB RAID drives that cost €800 for the same storage space. Getting to 6TB usable with RAIDZ2 would cost in excess of €1100 with Western Digital RE4 2TB drives. It would be pretty much cheaper operation to just get two sets of 1,5TB drives and mirror them. Except for the power usage and likely also noise.

If I get the software out and about within the next 6 months I might actually try to monetize it. I doubt there is a lot of money in it, though to my understanding the RTM guys do make a living with it. The price range for the extra features of Pro or Premium version is $3/month to $25/year. Even for a single developer you do need reasonable amount of subscribers to make a living. Anyways, we’ll see if anything comes out of this. There’s pretty much work involved in making everything ajaxy.

Still, this is interesting exercise, especially as both Django and Python are really new to me and so is the whole web framework world. Back when I worked on a .Net webapp using VB.Net pretty much everything had to be done by hand. Luckily these days I’m a lot better developer and much more interested in the whole field of sotfware development.

A comparative study was
performed to assess how professional programmers
use APIs with required parameters in objects’
constructors as opposed to parameterless “default”
constructors. It was hypothesized that required
parameters would create more usable and selfdocumenting
APIs by guiding programmers toward the
correct use of objects and preventing errors. However,
in the study, it was found that, contrary to
expectations, programmers strongly preferred and
were more effective with APIs that did not require
constructor parameters.

So essentially the explicit approach of

1
2
3

var foo = new FooClass();
foo.Bar = barValue;
foo.Use();

was better than the denser

1
2

var foo = new FooClass(barValue);
foo.Use();

(code examples straight from the article.)