Some additions to my holiday reading list:

• John Backus's Turing Award Lecture from 1977, Can Programming be Liberated from the Von Neumann Style? (via)
• Science is linguistic as well as numerical (just ask Larry Wall!)
• 10 Habits of Highly Effective Brains

I'm a little disappointed the last didn't explicitly require Chess since Kath gave me an awesome Kasparov Chess set for Christmas :)

Perl hacker Schlomi Fish recently published an article about the need for accessible Spam, "Spam for Everyone" - The International Campaign for Accessible Spam

The "Spam-for-Everyone" campaign aims to convince spammers that they shouldn't send spam that makes it harder or impossible for disabled people to understand it

Brilliant!

I have a few flights and trips over the Christmas week, so I thought I'd compile some reading material for those spare hours. If you're looking for some light ;) reading I thought you might be interested also:

• Gödel, Nagel, minds and machines
• Engineering Software Correctness
• User-Level Transactional Programming in Haskell
• Expression and Loop Libraries for High-Performance Code Synthesis
• It's Time to Stop Calling Circuits "Hardware"

Well it's the 20th anniversary of when Larry Wall first released Perl to the world and to celebrate the Perl community has announced the release of perl 5.10, the biggest new perl release in 5 years.

May we happily program so pragmatically for another 20 :)

Outside of work, SMS and instant messaging are fast becoming the writing tools of choice. Indeed, South Korea - that crystal ball of all our digital tomorrows - has even seen a report that many teenagers have stopped using email altogether. "It's for old people," they say.

...

According to the poll, mobile texting, instant messaging and the perception that email is "a lot of bother" are all contributing to the end of the email era. Other factors, say the report, are the difficulty of ascertaining if an email has arrived and the lack of immediate response. One young Korean said that texting felt like a ping-pong game and that email was more "like doing homework".

Source: Sydney Morning Herald

There's an opportunity in here somewhere - it shouldn't be too hard to merge IM, SMS and email into one seamless communication. Unfortunately Facebook/Cyworld etc. are always going to be a proprietary impediment.

While reading the erlang mailing list I came across a (fairly) new module plists containing parallel versions of the erlang lists module and a few throwins (like a mapreduce implementation). The functions operated just like the standard variants plus an optional parameter that allows you to tune how many parallel processes are run and whether to use remote nodes or not.

plists is a drop-in replacement for the Erlang module lists, making most list operations parallel. It can operate on each element in parallel, for IO-bound operations, on sublists in parallel, for taking advantage of multi-core machines with CPU-bound operations, and across erlang nodes, for parallizing inside a cluster. It handles errors and node failures. It can be configured, tuned, and tweaked to get optimal performance while minimizing overhead.

...

This module also include a simple mapreduce implementation, and the function runmany. All the other functions are implemented with runmany, which is as a generalization of parallel list operations.

Nice.

I wanted to serve compressed static js and css content from an Apache server. I also wanted to pre-compress the files to save time/cpu, and the particular Apache 2 instance had no mod_deflate or mod_gzip installed (dynamic content was being compressed at the application layer).

Serving pre-compressed content turned out to be incredibly easy using mod_rewrite*:

1. Create a compressed copy of each .js & .css file with a .jsgz/.cssgz extension
   

   find . -name '*.js' -or -name '*.css' -exec sh -c 'gzip -c {} > {}gz' \;
   

2. Add the following to a relevant part of your Apache config file:
   

   RewriteEngine on
   RewriteCond %{HTTP:Accept-Encoding} gzip
   RewriteRule (.*)\.css$ $1\.cssgz [L]
   
   AddType "text/css;charset=UTF-8" .cssgz
   AddEncoding gzip .cssgz
   
   RewriteCond %{HTTP:Accept-Encoding} gzip
   RewriteRule (.*)\.js$ $1\.jsgz [L]
   
   AddType "text/javascript;charset=UTF-8" .jsgz
   AddEncoding gzip .jsgz

3. Test & reload your apache config
   

   httpd -t
   httpd -k graceful

Voila, correctly served pre-compressed static files. Just be sure to update your compressed versions whenever you update the content.

I took part in my first ever podcast last week on a panel discussing the Australian IT job market. Mark Jones hosts the weekly MIS Australia podcast for the AFR and invited me to represent the IT candidate market.

Also on the panel was Jane Bianchini, COO of Candle ICT (which seems to have changed it's name to Clarius in the last week - Moof ;), Carl Sjogreen, senior product manager at Google Australia and Peter Sheahan, an entrepreneur and author about Gen Y.

It was a good experience - I did just manage to get a few words in edgewise!

Use the embedded player or why not visit The Scoop and enjoy weekly panels talking about issues affecting the Australian IT industry.

Of course super-light threads for parallel programming is Erlang's killer feature, so let's apply that to a simple merge sort. Erlang's spawning and message passing interface is fairly straightforward, but a lot of parallel tasks are simple start -> process -> return type tasks, so we can simplify our syntax even further by using some nice helper functions such as Joe Armstrong's parallel map pmap or Anonymous Matthew's parallel eval pareval (via Philip Robinson's blog).

Parallel map makes light work of a massively parallel merge sort:

-module(mergesort).
-export([msort/1]).

%% terminal cases for list of less than 2 entries

msort([]) ->
    [];
msort([A]) ->
    [A];

% split list into 2 & mergesort them

msort(L) ->
    Mid = trunc(length(L) / 2),
    {A, B} = lists:split(Mid, L),
    lists:umerge( pmap(fun(I) -> msort(I) end, [A, B]) ).

This gets really interesting in certain cases eg. where acquisition of the data is expensive but randomly accessible.

Instead of passing in a simple list, our sort function could accept a future that would promise to return any requested linear subset of the data. Assuming we know the total length of the source list, no data is needed until we reach the merge phase at the bottom of the recursion stack, and incrementally more data is required as we bubble back up to the final result.

In this example we are doing our network retrieval in parallel with our sorting (which itself is happening in parallel). If the remote data comes from multiple hosts we could parallelise the retrieval by using multiple port instances.

Like every man and his dog I've been playing around with erlang (and a smattering of Haskell for good measure :). I thought I'd try to see how I might write a binary tree implementation Erlang.

I must confess it turned out much less elegant than I thought it might - maybe some erlang or other functional gurus can point out a better algorithm.

Now before you all comment, yes I know most of the below isn't tail recursive - but performance isn't really the issue here, and a binary tree implementation in erlang is going to be inefficient anyway if you're making frequent updates, since every recursion will involve copying some of the structure (maybe it optimises that away ... dunno enough about its internals to comment).

-module(tree).
-export([find/2, tree_of_squares/1, insert/3]).

-record(node, {key, val, left = nil, right = nil, height = 1}).

%% find/1

find(SearchKey, Node) when SearchKey == Node#node.key ->
  {ok, Node#node.val};

find(_, nil) ->
       not_found;

find(SearchKey, Node) when SearchKey < Node#node.key ->
  find(SearchKey, Node#node.left);

find(SearchKey, Node) ->
  find(SearchKey, Node#node.right).


%% tree_of_squares/1 : make a tree where key = val and contains
%% squares of all numbers from 1 to Max


tree_of_squares(Max) ->
  tree_of_squares(Max, nil).


tree_of_squares(0, Tree) ->
  Tree;

tree_of_squares(Max, Tree) ->
  insert(Max * Max, Max * Max, tree_of_squares(Max - 1, Tree)).

%% insert/3 : return a new tree made from inserting (or updating)
%% a node of Key, Val into existing tree or new tree (nil)

insert(Key, Val, nil) ->
       #node{ key = Key, val = Val };

insert(Key, Val, Node) when Key == Node#node.key ->
       Node#node{ val = Val };

insert(Key, Val, Node) when Key < Node#node.key ->
       balance(Node#node{ left = insert(Key, Val, Node#node.left)});

insert(Key, Val, Node) ->
       balance(Node#node{ right = insert(Key, Val, Node#node.right)}).

%% height/1 : return the height of the tree

height(nil) ->
       0;
height(Node) when Node#node.left == nil, Node#node.right == nil ->
       1;

height(Node) ->
       height(1, Node#node.left, Node#node.right).

height(Acc, Left, Right)
 when Left == nil;
      Right#node.height > Left#node.height ->
       Acc + Right#node.height;

height(Acc, Left, _) ->
       Acc + Left#node.height.


%% balance/1 : return a new tree which is the balanced version of Tree
%% note also that balance is trusted to always return the root height set
%% correctly

balance(nil) ->
  nil;

balance(Node) when Node#node.left == nil, Node#node.right == nil ->
       Node#node{ height = height(Node)};

balance(#node{ left = L, right = R} = Node) when L#node.height ==
R#node.height ->
       Node#node{ height = height(1, L, R)};

balance(#node{ left = L, right = R} = Node)
 when R == nil ;
      L#node.height + 1 > R#node.height ->
       swing_right(Node);


balance(#node{ left = L, right = R} = Node)
 when L == nil ;
      R#node.height + 1 > L#node.height ->
  swing_left(Node).

swing_left(#node{right = #node{ left = RL, right = RR } = Right} = Node)

 when RR == nil ; RR == nil, RL == nil ;
      RL#node.height > RR#node.height ->

       move_left(Node#node{right = move_right(Right)});

swing_left(Tree) ->
       move_left(Tree).

swing_right(#node{left = #node{ left = LL, right = LR} = Left} = Node)

 when LL == nil ; LL == nil, LR == nil ;
      LR#node.height > LL#node.height ->

       move_right(Node#node{ left = move_left(Left)});

swing_right(Tree) ->
       move_right(Tree).


move_left(Node)
 when Node == nil ;
      Node#node.left == nil, Node#node.right == nil ->
       Node;

move_left(#node{ right = #node{ left = RL } = Right} = Node) ->

       OldRoot = Node#node{ right = RL },
       NewRoot = Right#node{ left = OldRoot#node{ height = height(OldRoot)}},
       NewRoot#node{ height = height(NewRoot) }.


move_right(Node)
 when Node == nil ;
      Node#node.left == nil, Node#node.right == nil ->
       Node;

move_right(#node{ left = #node{ right = LR } = Left } = Node) ->

       OldRoot = Node#node{ left = LR},
       NewRoot = Left#node{ right = OldRoot#node{ height = height(OldRoot)}},
       NewRoot#node{ height = height(NewRoot) }.

tree_of_squares is just to easily create an interesting tree to play with, for instance:

1> c(tree).
{ok,tree}
2> T = tree:tree_of_squares(7).
{node,25,
      25,
      {node,9,
            9,
            {node,4,4,{node,1,1,nil,nil,1},nil,2},
            {node,16,16,nil,nil,1},
            3},
      {node,36,36,nil,{node,49,49,nil,nil,1},2},
      4}

Then you can try to find a node:

3> tree:find(1, T).
{ok,1}
4> tree:find(2, T).
not_found

Not very interesting when the key = val - let's insert and update some entries:

5> T2 = tree:insert(5, bar, tree:insert(4, foo, T)).
{node,25,
      25,
      {node,9,
            9,
            {node,4,foo,{node,1,1,nil,nil,1},{node,5,bar,nil,nil,1},2},
            {node,16,16,nil,nil,1},
            3},
      {node,36,36,nil,{node,49,49,nil,nil,1},2},
      4}
6> tree:find(5, T2).
{ok,bar}

Fun huh!

An implementation of red-black trees is left as an exercise for the reader ;)

Computer Science is no more about computers than astronomy is about telescopes.

--E. W. Dijkstra

Great quote from the introduction of Algorithms with Perl

While stuck in Apple-reminiscing mode (I'm reading Revolution in the Valley) I thought I'd see if someone in Apple was cool enough to have a server running on bric-a-brac.apple.com.

Unfortunately it seems noone at Apple is nearly that cool:

Failed to resolve the name of server bric-a-brac.apple.com to connect

I love a good Round Rect as much as the next guy (see Round Rects Are Everywhere) and I am really happy about menus being rounded in Leopard. I dunno what it is, it just makes me feel nice every time I drop open a menu!

I was browsing around the various google sites tonight to see what had changed, and I thought I'd punch my name into google code, the search index for source code.

There are surprisingly few results, but even worse is that the second search result is:

So any prospective collaborator or employer discovers that I have an un-done todo dating back to March 5 2003! And there's not even any comments suggesting what needs fixing. Oh well, it's in use in hundreds of production systems with tens of thousands of users, so I guess it doesn't really need fixing after all!

A while back I found this neat sed pipeline for extracting an ini file section into shell script variables:

INI_FILE=path/to/file.ini
INI_SECTION=TheSection

eval `sed -e 's/[[:space:]]*\=[[:space:]]*/=/g' \
    -e 's/;.*$//' \
    -e 's/[[:space:]]*$//' \
    -e 's/^[[:space:]]*//' \
    -e "s/^\(.*\)=\([^\"']*\)$/\1=\"\2\"/" \
   < $INI_FILE \
    | sed -n -e "/^\[$INI_SECTION\]/,/^\s*\[/{/^[^;].*\=.*/p;}"`

You now have variables in the scope of your shell script derived from the key/value pairs in that section of the ini script. REALLY handy. Thought it might help someone else if I recorded it here. I got it from this thread on debian-administration.org.

NB: the eval is a ksh builtin. I imagine using declare in bash will achieve the same, but I haven't tested it.

NB2: it seems Solaris sed isn't quite up to the task - the above requires gnu sed.

I've come across a few more (very minor) Leopard glitches.

The first is that if you have a Finder window with no horizontal scrollbar, and you then click on a file that's right at the bottom of a file list and which, by displaying it's preview, forces a horizontal scrollbar, then the file list doesn't get pushed up. That results in the image on the right - notice how you can just see the blue highlight of the X11 file poking above the horizontal scroll bar?

The second is with the dock and X11 (not sure if it's an issue with the app or the dock). If you put X11 in the dock, and then start it (from the dock), the icon you put in the dock bounces forever and a new X11 icon pops on the bottom of the dock (like it does for apps that are not in the dock already). you can see above I even sometimes get two extra icons. Note that I'm not starting any additional instances of X11 or any X apps besides the initial default xterm.

Update: The multiple X11 icons also appear in the Command-Tab bezel.

Update 2: It seems there are some pretty massive changes to X11 in Leopard. For the better Apple has switched to X.org, but that (and other) changes have been major and some bugs have slipped through. The change to launchd is interesting - while non-intuitive (in an X11 way) it makes X11 apps more intertwined with the main OS.

Anyhoo, thankfully the X11.app maintainer is on the ball and has posted an update to solve the two most glaring issues - the faulty three-button emulation (critical on a laptop) and multi-monitor support. Go to http://www.x.org/wiki/XDarwin for more. I still get some really wacky behaviour when my monitor resizes (eg. by plugging my 22" crt into my 12" powerbook) but it's better.

I've found what seems to be a recurrant glitch on my new Leopard install - disk images just don't unmount via the Finder. Download and mount any .dmg and no matter how many times you drag them to the trash, click the eject icon in the sidebar or right click->eject, they stubbornly stay mounted.

It seems to be a Finder issue rather than with the disk image driver since a simple umount /Volumes/DiskImageName makes them disappear pronto.

Update: Further, Finder also gets terribly confused in some circumstances where you mount two (different) images with the same volume info. I got into one state where there was one volume showing in the sidebar that wasn't even mounted. A swift relaunch of the Finder sorted that out...

Update 2: Seems after relaunching the finder the original issue itself went away.

So I easily got an Amazon S3 storage bucket mounted under Leopard, but it turns out Time Machine doesn't support network mounts.

Well it's not that it can't, it actually does - but it won't show you any disks that it doesn't think are local (eg. disks served by Airport Express apparently work fine), so Apple evidently have decided that my net connection isn't fast enough for backups so I'm best off not even trying :(

So let's start hacking :) The TimeMachine preference pane binary has the following methods:

@interface AppleTMPref (Private)
- (BOOL)_isSupportedNetworkMountPoint:(id)fp8;

@interface AppleTMSettings : NSObject
{
    ...
}
- (BOOL)isNetworkBackup;

@interface TMBonjourBrowser : NSObject
{
    NSNetServiceBrowser *_browser;
    NSMutableArray *_netServices;
    NSImage *_serverImage;
}

- (id)init;
- (void)dealloc;
- (id)_dictionaryFromAirPortDiskData:(id)fp8;
- (BOOL)_isLocalNetService:(id)fp8;
- (id)_volumesFromNetService:(id)fp8;
- (id)bonjourDisks;
- (void)netServiceBrowser:(id)fp8 didFindService:(id)fp12 moreComing:(BOOL)fp16;
- (void)netServiceBrowser:(id)fp8 didRemoveService:(id)fp12 moreComing:(BOOL)fp16;
- (void)netServiceDidResolveAddress:(id)fp8;
- (void)netService:(id)fp8 didNotResolve:(id)fp12;

@end

@interface SnapshotUtilities : NSObject
{
}

+ (BOOL)mountPointIsNetworkVolume:(id)fp8;

Time to crack out F-Script...

Ok so I'll write this up better later, but here are the steps I followed to successfully mount an Amazon's S3 network storage bucket under Leopard:

1. create an amazon web services account
2. sign up for amazon S3
3. download MacOS X S3 Browser
4. install MacFUSE Core for OSX 10.5
5. get your Amazon S3 key ID & secret (you'll get an email from Amazon with an url about that after signing up)
6. login and create an Amazon S3 bucket using the S3 Browser you downloaded (I called my bucket "12inch-timemachine" for backing up my 12" Powerbook)
7. download s3fs.cpp from the s3fs project
8. you'll need to add the following line to the top of the file somewhere:
   

   #define __off_t off_t

   (props to Roberto Saccon for figuring this out)
9. compile s3fs.cpp with the following incantation:
   
   g++ -Wall -D__FreeBSD__=10 -D_FILE_OFFSET_BITS=64 -I/usr/local/include/fuse -L/usr/local/lib -lfuse -pthread -lcurl -lcrypto `xml2-config --cflags --libs` -ggdb s3fs.cpp -o s3fs
10. create the file /etc/passwd-s3fs with your amazon s3 key id and secret separated by a colon ie:
    example-id:example-secret-key (note that you'll want to be careful about the permissions on this file)
11. mount the S3 bucket as an OSX directory like so eg:
    

    mkdir ~/s3
    ./s3fs your-bucket-name ~/s3 -ovolname=Flubber

    where Flubber can be whatever you want the mounted volume to appear named as
12. you can now access the mounted volume at your mount point (in the above example ~/s3) or as a Finder Volume which you can see by going to "Computer" in the Finder's "Go" menu. Not sure why it doesn't show up in the Sidebar)

Update: It seems Leopard, by default, doesn't show network volumes on the desktop by default. You can turn that on in finder preferences, but your volume still won't show up in the sidebar (under either Devices or Shared). A simple solution is to us the local option like so:

./s3fs your-bucket-name ~/s3 -olocal,volname=Flubber

Unfortunately that still doesn't allow it to work with time machine. You also get a folder icon on the desktop instead of a disk icon, but it's in the sidebar under Devices now. The icon even has a working eject icon which is handy, but it leaves an annoying folder icon in the sidebar.

Update 2: the ping_diskarb option seems to help keeping the Devices icon current:

./s3fs your-bucket-name ~/s3 -olocal,ping_diskarb,volname=Flubber

Update 3: The previous must be in my imagination:

On Mac OS X 10.5.x, both these options [ping_diskarb, noping_diskarb] are no-ops

I don't remember hearing about this, but default Cocoa text shortcuts (eg. emacs keys like Ctrl-a for home, Ctrl-e for end) work in Finder now.

Well I rushed out and bought Leopard, mostly for the garbage collected goodness of Objective-C 2.0.

The look and feel changes are mostly nice, but I have found one minor tweak that irritates me no end - and doesn't make sense to me from an HCI point of view.

If you have opened a drop down menu, and you decide you are happy with the existing selection, it used to be that you could "dismiss" the menu and leave it as it is by clicking anywhere on the screen outside the menu.

Now you need to either click on the selected option to re-select it, or if there happen to be unavailable options in the menu (ie. greyed out) you can click them as well.

Now that sucks from an HCI standpoint in two ways. First, to execute this relatively common action you now need to click in a quite small area instead of simply flicking your wrist and clicking. Secondly it is inconsistent with the menus in the menu bar, where you can click anywhere outside an open menu to dismiss it.

Update: It seems to have been another "glitch" because after a reboot I can no longer replicate it. Little bugs happen - at least it's not a change in the ui thinking.

The perfect outcome I was daring to hope for!! My prediction was for Lewis and Alonso to take each other out on the first corner, but it seems somehow fitting for Alonso to miss out on the championship by simply fading away.

Kimi really deserves this championship. He's driven fantastically all season in what was a new car and new tyre for him. After years of unreliability with McLaren it must feel all the sweeter.

Well done Kimi, great job. Have a vodka for me :)

Roberto Saccon has successfully used a FUSE based S3 driver to mount an Amazon S3 bucket as a regular Mac volume.

That + time machine in the upcoming MacOS X 10.5 will make for a very cost effective and scalable backup solution.

Sweet!

Mark Webber qualifying 5th at Inter Lagos - brilliant! Massa on pole - even better (although I would have preferred to see Kimi in P2).

By the way sorry everyone for the dearth in technical posts - been a bit busy lately with a few things (including working on my lap time down at Wakefield park - 1:24 is my best so far :).

I said at the start of the race that Mark Webber would be great in the wet.

Even the ITV boys thought he had a chance at Lewis for a win once the safety car period ended. If not, a dead certain second place - bettering his previous best of third.

Even throwing up in his own helmet didn't stop him.

But Sebastian Vettel smashing into him did.

Heartbreaking.

Pity that Webber couldn't get a 10 place boost instead of Vettel recieving a 10 place penalty for the next race.

Semi from TramTown may not like Ratatouille (the new Pixar flick), but I loved it. 7/10

Nation leaders representing 56% of the world's GDP are in Sydney, and The Sydney Morning Herald website leads with "Code crack opens the iPhone floodgates"??

According to "one member of a group of Australians", "there are at least 30 or 40 people in Australia" using iPhones.

Using phrases like "thumbing their noses" and "much-ballyhooed", is this really Australia's premier journalistic talent? This could be an article straight from The Onion (which always has had fancier prose that it's Australian equivalent The Chaser) but unfortunately it's deadly serious.

Maybe I'll have to turn to mX for my serious editorial content.

I've just finished reading an interview by Grady Booch (a pre-eminent computer scientist himself) interviewing Andy Hetzfeld and Bill Atkinson titled The Oral History of MacPaint.

Bill Atkinson is probably one of my biggest heroes of software design (Steve Wozniak and Burrell Smith being my heroes of hardware). Another one of my heroes of software (or really computing) design is Douglas Engelbart. In my mind the time frame from one to the other was huge, but it's really only 15 years or so - and Bill met Douglas Engelbart - astonishing!

So, I said at '68, the windows and mouse were invented by Englebert, and I actually went to visit Englebart. It was a wonderful visit, and I think he gave a really good argument which was that when you design for the beginner, sometimes you shortchange the experienced user. If you were designing only for a walk-up-and-use-it experience, you would never design a bicycle because it takes a while to learn to use a bicycle, although it gives you good power. So, he was designing more for a professional knowledge worker, and he's willing for people to learn this five-key chord thing and lots of new ways of working with stuff.

Now very few of the facts in the interview were news to me like that, but discussed as such a compact timeline of the birth of the GUI and desktop application as we know it (taking in the foundational elements of Engelbart's NLS and Xerox Parc's Smalltalk) makes a number of things really stand out. Those notes are below, along with some nice little one liners I have quoted or paraphrased.

1. The entire design of what is used productively today on every PC in the world was designed by technology people. I mean *really* technical people, but people who also really cared about the purpose - how it would help peoples' lives. Today many environments don't allow their technology people to design a newsletter without some sort of external designer or consultant, and by organising this way we discourage technology people from becoming engaged and involved - instead restricting themselves to mechanical implementation.

2. Sometimes you do a lot of work just to throw it away - but the later things you do wouldn't have happened any other way. And you never know which is going to be the important thing that flops, the thing that just flops, or the thing that succeeds - so you have to put your all into it anyway. The Lisa was a total flop. Everyone knows that. But Apple spent a lot of time investing in the Lisa product, and the legacies of that were Bill Atkinson's (and other's) familiarity with the 68000, the foundations of QuickDraw, the testbed hardware for the mouse driven GUI, etc.

3. You're never too busy for a fun outing. Imagine if Steve Jobs et al had been "too busy" to take a tour around Xerox Parc. They would be forgiven for saying their filofax was full.

4. Your biggest competitor is not a big corporate, it's a three person research team

5. Computer processing cycles aren't valuable. People's time is valuable.

6. If you find yourself thinking "what would anyone use that for?" think harder ("Apple was offered lock, stock and barrel exclusive
   rights on a spreadsheet for $50,000, and they turned it down. ... What would anyone use it for? They could have owned the whole spreadsheet concept.")

7. Sometimes the crazy genius is right instead of the conservative committee.

8. Sometimes your boss is wrong and he'll thank you later for ignoring him (Steve Jobs insisted that "the Mac would only be 128K ever, and that it was not okay to have it expandable to 512. Actually, Burrell went behind Steve's back to make it so there was one trace you could cut, and you could stick in the 512K chips. Steve later had to apologize and say he was wrong. ")

9. Think about things a lot. Try some things. Try different things.

10. Some mistakes you make will stick around for a very long time (Atkinson: Well, a design I regret is the double click.)

11. Sometimes a non-optimal solution is the best one.

12. Brain surgeons are really smart.

13. Code should be beautiful (Atkinson: I would spend time rewriting whole sections of
    code to make them more cleanly organized, more clear.)

14. Code comments shouldn't have to say what something does - instead should say why it is being done.

15. Employ great graphic designers and writers.

16. Having to teach somebody how a piece of software works forces you to clean it up.

17. Sometimes you make a product better by leaving features out.

18. technology is very cool, and
    there's beauty in the software, but software is ultimately a human experience.

19. If you want to get it smooth, you've got to rewrite it from scratch at least five times.
    And have a lot of user testing. Because you can't see the things that you can't see.

Footnote: It's interesting that Grady starts using the word 'heroes' part way through the interview, because I had already come up with the title for this blog before I got that far.

86 Mac Plus Vs. 07 AMD DualCore. You Won't Believe Who Wins

Classic - literally.

For the functions that people use most often, the 1986 vintage Mac Plus beats the 2007 AMD Athlon 64 X2 4800+: 9 tests to 8! Out of the 17 tests, the antique Mac won 53% of the time! Including a jaw-dropping 52 second whipping of the AMD from the time the Power button is pushed to the time the Desktop is up and useable.

...

Is this to say that the Mac Plus is a better computer than the AMD? Of course not. The technological advancements of 21 years have placed modern PCs in a completely different league of varied capacities. But the "User Experience" has not changed much in two decades. Due to bloated code that has to incorporate hundreds of functions that average users don't even know exist, let alone ever utilize, the software companies have weighed down our PCs to effectively neutralize their vast speed advantages. When we compare strictly common, everyday, basic user tasks between the Mac Plus and the AMD we find remarkable similarities in overall speed, thus it can be stated that for the majority of simple office uses, the massive advances in technology in the past two decades have brought zero advance in productivity.

I wouldn't say zero - the pivot table alone has improved savvy spreadsheet users efficiency a lot, and they didn't mention that Word 3 didn't have any real table support. But that's not really the point. It's still outrageous how little improvement the average user of office productivity software has gained from the billions expended on so-called software development.

I have commented before about interesting search terms I see in my web server stats. You'll be pleased to know that Mel Kaye and magic mug cake are still frequent contenders for the top search terms.

I noticed a disturbing search term this month though:

evolution of programming languages in power point

What are they thinking??!

Of course should they be interested in programming languages they will find a feast in this blog: perl, SQL, ruby, emacs elisp, smalltalk, java, C, 6502, arm,NLS, Cocoa/Objective-C, future issues for theoretical computer science and even visual basic! I'm sure there's more there, and erlang and haskell are sure to follow...

But *never* will there be a good thing to be said about powerpoint on this blog :)

People who know me know all about switching all of my home entertainment off discrete units to a spare Powerbook. I tell anyone who will listen how great it is!

All you need is a mac, some powered speakers, and a screen. Now this is nothing new - people have been doing it with Mac Minis ever since they were released. I prefer a laptop since some things are just easier to do with a keyboard and mouse - and I can do them without stopping playback on the main display.

One thing that was missing was unifying control - by that I mean sitting on my couch, using one remote to do everything without getting up. What point is home entertainment if you can't drive it from your couch?!

There's a bunch of projects that aim to do this, but I wanted something simple. No on-screen menu, just the standard features from a remote.

So I bought the excellent Keyspan Express remote. I would have preferred an RF remote (rather than infra-red), but the affordable RF models have less buttons.

Then the challenge was to use the limited number of buttons to achieve an understandable (to non-technical people) set of functions that would let you do most things.

Over some coming posts I'll explain what actions I chose for what reasons, but the curious can take a look at some applescripts in the mean time.

1 degree - A News Limited Initiative

Everyone's favourite media mogul is turning down the thermostat 1 degree, which will save 5% on heating usage.

http://www.1degree.net.au/

The video is worth watching. Even Rupert realises that acting on climate change is "simply good business". No overt attempt to pretend a moral high ground.

Carbon neutral by 2010 - tough ask for a company with a carbon footprint of 641,150 tonnes, but Rupert isn't known for missing his targets.

And perhaps this is the real momentum building. While democratic government waits for an overwhelming majority others will step in. Global business, aid groups, the courts, churches. It gets interesting from here.

I seem to have been Listed in blogs.com.au - whatever that is.

Only Australia could turn out blogs named Frugal Bastard and elliot's artwank :)

Ross Dawson discusses an SMH article which quotes his discussion (still following?) of the state of blogging in Australia, tying it in with bandwidth:

"Of the top 25,000 blogs globally, around 9000 are in English", says Mr Dawson. Of those, only 75 originate in Australia. But there are 420 million native English speakers in the world. "With Australia's population of 21 million, we comprise 5 per cent of English speakers. But with 75 blogs out of 9000, we comprise less than 1 per cent of English blogs. We are underrepresented by a factor of six or so."

...

"I think one key reason is lack of bandwidth in Australia, and of its high cost. Australian internet connections are slower than they are in the rest of the world, and Australia is almost unique in capping usage at quite low levels."

There should be some adjustment in interpreting those numbers for local vs. global readers.

You could only make such a simplistic percentage comparison for countries of a similar size.

For instance a widely read blog on local Californian politics will probably rate in the top 25,000 blogs globally - since California has such a huge population. A widely read blog on local NSW politics could easily have a higher percentage of the available readership, be more influential etc. - and yet not rate a mention in the top 25,000 blogs.

Now I'm not saying that the general observation is wrong - but these figures don't back it up as clearly as it may first appear.

As for bandwidth - blogging doesn't take a lot of bandwidth. Price is a far more important factor. Very cheap always on internet is going to promote more blogging than very high bandwidth internet that doesn't reduce the cost of the cheapest plans.

But even with cost, I don't think that the key issues are so simple. As a nation we seem to be underrepresented in entrepreneurship in general, and I think blogging is simply literary entrepreneurship.

Solving that chestnut is a rather difficult socio/economic/politic/educational issue - beyond the scope of this Australian blogger at the moment ;)

Original linker: Corporate Engagement.

I just spent a few evenings in autohell trying to configure the cvs trunk of GNU Smalltalk.

I wish I could document my story, but there were just too many twists and turns (many which ended with the distinct feeling of darkness and that one was likely to be eaten by a grue ;).

Instead I will just document some things I found useful. Firstly, install the latest autoconf and automake package your packaging tool has (MacPorts in my case). Secondly, consider the following your personal arsenal:

• autoupdate --verbose
• find . -name configure.ac | xargs perl -pie 's/AM_CONFIG_HEADERS/AC_CONFIG_HEADERS/'
• autoreconf --verbose

I had a need to alter a bunch of ids and names after I cloned a dom node, so I cooked up the following dom node visitor:

function domVisitor(el, f) {
  if (el) {
    f(el);
    for (var i=0 ; i < el.childNodes.length ; i++) {
      domVisitor(el.childNodes[i], f);
    }
  }
}

So all you need to do is pass in the cloned node (or whatever) along with an anonymous function to do your bidding. The function will be called for the full tree under the node passed in.

Groklaw posts the transcript of an excellent speech by Eben Moglen, one of the co-authors of the GPL v3 open source software license, a Professor at Columbia Law School in New York and the founder of the Software Freedom Law Center.

He begins with a surprisingly striking analogy:

I ask you to imagine briefly a world in which arithmetic has become property. ... Once we have reduced arithmetic to property, you'll have only as much arithmetic as you can afford, the consequence of which is that the gateways to material collaboration in the world, successful activity in relation to the physical and constructed environment, will depend very largely upon one's ability to acquire sufficient surplus amounts of mathematics.

Then invoking the famous words of Richard Stallman,

Why is software property? It should be knowledge to be shared, like math, like physics. It's unethical, to deprive people of information evidently available to them about the artefacts of digital society with which they are daily in contact -- it's evidently immoral to deprive them of knowledge. You've given the knowledge to the computer sitting next to them. They're using it -- the knowledge is playing a potentially determinative role in their lives. You've already delivered it to them -- all you haven't done is to deliver them the ability to know.

and also paraphrased John Dewey (who I had never heard of):

the education and expansion of the human mind depends upon the opportunity to experiment with the world.

talks quickly about the history of copyright in the US, and then turns to the new license itself.

He also has an interesting discussion comparing the pragmatic commercial foundations of US law and UK/Commonwealth common lay with what is happening online today:

But I would present to you the possibility that the UCC and the GPL 3 are in themselves a pair -- a pair, organizing an idea about the method of the creation of 21st century law. 21st century law is born in the street in the same way 21st century television is born in the street, not sent to you from the top of a broadcast tower, but upward from the cellphone and the portable camera put through YouTube.

21st century law is like 21st century music -- not made in an expensive recording studio or legislature, rented by the hour by people with the power to rent studios and legislatures, but made in every laptop in every den in every corner living room, in every garage, where a musician and a computer are, which is pretty much everywhere a musician is.

...

Of course, the idea that law might be something we all do together, it's got a long history stretching back far beyond the Free Software movement, stretching, in fact, back to beyond democracy. One of the characteristics that the continental Europeans noted of the English speakers, North Britains and South Britains, similarly in the course of the 16th, 17th, and 18th centuries, was that English-speaking people had an almost personal relationship to the common law. A man might be an artisan or a yeoman farmer, but he believed the law to be in some sense his own. He was familiar with the courts, he served on juries, the language of the law was in his mouth. Even beyond the language of literature and religion, it was in his mouth. It was, if not folk law in some forests-of-Germany sense, community law. The law of us, and to be not of this law was to be not of us in some fundamental way.

This is a must read for anyone technically or legally inclined or engaged in 21st century society.

I wonder if there is anyone in the world who has access to a tape measure who has *not* done this:

 

 

Source: xkcd - the best nerdy cartoon in the world.

• Have some tabs open;
• type something into a text box in a web page;
• close that tab, but of course the window is still open because of the other tabs
• Edit->Undo is still available since the Cocoa UndoManager works, by default, per window;
• Undo ->Crash :)

That's something to watch out for when you break the window instance == document instance that Cocoa assumes.

Note that this is with the Safari 3 beta - but I'd be willing to bet the same behaviour occurs in Safari 2. Can anyone confirm/deny?

When it comes to switching between objects and natives, Perl OO is not quite as orthogonal (or DWIM) as I would like. For instance, assuming I have a class:

package MyArray;

...

sub push { ...}

Then I can treat an instance of that object as an object with a method push, or as a kindof array (kindof, because of course an actual array would have @ for its sigil):

# these are equivalent:

$instance->push(1);

push $instance, 1;

I don't, however, have the orthogonal behaviour with an actual array ref:

# this works:

push @$arrayref, 1;

# this doesn't:

$arrayref->push(1);

# nor, of course, does this:

@$arrayref->push(1);

What I would like is more DWIM than real orthogonality, but if Perl's not DWIM then it's nothing! I guess I'm just wishing for real native array objects like Ruby (sigh).

Dan Bricklin (co-creator of VisiCalc) has always been a hero of mine. So many assumptions of current software come from his simple innovations. His key focus was simplicity and ease of use. I love this quote from one of his business school professors, giving him feedback on his prototype:

You're competing against the back of an envelope. It's got to be really easy to use.

-- Source: Founders at Work

What is your product competing with?

Better than receiving an sms, GooSync bridges between my google calendar and SyncML which means I can bi-directionally sync it with the calendar built into my SonyEricsson phone.

It works pretty well. They support Google's new auth method that means you don't need GooSync to store your details. Unfortunately they don't have that working with Google Apps domains (which I use), so I have to use their other solution which involves them storing your details in their database (apparently encrypted - although obviously reversible somehow... I think they encrypt it with your GooSync password).

The auto-config sms didn't seem to work (I received the settings, but they failed to apply on my Z800i) but it was easy to manually enter the appropriate settings. Then the first sync failed, but all subsequent sync's have worked fine. I've set my phone to sync once a day so I'll see how that goes and report back on the bandwidth used.

Thomas Enebo has just posted JRuby 1.0.0RC3 which sounds like the final release candidate before 1.0.

This will likely be our final release candidate before our 1.0 release. People are encouraged to try out this release to help us find any remaining showstopper issues. We have spent a lot of time over the last month squashing compatibility bugs and we have confidence that applications 'will just work' (tm)*. Please try your applications and libraries against JRuby and give us feedback.

Sources: ruby-dev mailing list, Tom's Ruminations.

I never thought I would get so excited about a technology that's even remotely Java-related ;)

The BBC has a great article on F1s professional safety car drivers.

"It looks boring on the TV when the safety car comes out" Maylander tells BBC Sport. "It looks really, really slow, I know."

The thing is, it isn't.

Maylander's safety car is a Mercedes CLK 63 AMG, capable of reaching 62mph in 4.5 seconds (about twice the time it takes an F1 car), with modified brakes, tyres and aerodynamic parts; Maylander fondly describes it as "a small race car".

"The top speed I can reach, on Silverstone's long straight for example, is 155mph - but it looks like 60mph compared to the F1 cars," the German said.

It sounds pretty fun (and hair-raising), but they have so sit in the car, harnessed in, helmet on, for the whole race. And they need to be damn good drivers too - as Bernd Maylander (one of the two drivers) said, they would look pretty stupid if they went off!

Rob Tannen writes an excellent article on boxes and arrows titled "Simplicity: The Distribution of Complexity".

So, given a fixed set of elements (e.g. navigation choices), there are multiple attributes of simplicity than can be addressed to varying degrees. The success or failure of a design is largely dependent on achieving the right balance among these attributes of simplicity. An infamous example of this was BMW's i-Drive controller, designed to operate a range of automotive systems from a single control. The resulting user experience was confusing and complex as BMW relied too much on mechanical simplicity (a single, clickable rotary control) at the cost of perceptual and cognitive simplicity.

It's an excellent article, go check it out.

Software developers have gotten away with being sloppy because of the abundance and rate of growth of compute cycles. However, the problems of large numbers start to intrude: just as power and heat become factors in the presence of tens of millions of transistors, power and heat become architectural considerations in the presence of teraflops of computation and petraflops of storage.

-- Grady Booch, Power As A Limiting Factor

Mel Riffe reckon's he's a gardener or garbo*. What about you?

* Garbo is Aussie slang for a garbage/rubbish/trash/waste collector

It was only last year, as I was playing around with Smalltalk that I discovered the joy of an object database.

Naturally as I began playing with JRuby I wondered about object database options for it. One of the great things about the Java linkup is the wealth of available libraries and products and sure enough I found a number of open source OO database options. GOODS (which I have used with Smalltalk) is an option, but like some others it relies on a bytecode pre-processor, which makes me nervous given how JRuby plays with the bytecode.

Playing with db4o (which has both GPL and commercial licenses) has been pretty successful for the very little time I have invested.

Unsurprisingly I discovered that trying to store a pure Ruby object yielded an ObjectNotStorableException exception. In JRuby extending Java classes is easy. I haven't discovered how to properly extend the constructor, so I ended up with the following class:

require 'java'

class Song < java.lang.Object
  attr_writer :name, :artist, :duration
  
  def to_s
    "Song: #@name--#@artist (#@duration)"
  end
end

Creating a song object and storing it in a db4o database is simple:

include_class 'com.db4o.Db4o'

db = Db4o.openFile("/tmp/test_db")

song = Song.new
song.name = "Bicylops"
song.artist = "Fleck"
song.duration = 260
puts song

db.set(song)

I can then pull song objects out of the database using QBE:

# make an unconstrained template object for QBE
song_proto = Song.new
song_proto.name = ""
song_proto.artist = ""
song_proto.duration = 0

results = db.get(song_proto)

results.each {|s| puts s}

Unfortunately I don't get a good look at the objects since they appear as a generic proxy object. to_s yields:

(G) org.jruby.javasupport.proxy.gen.Object$Proxy0

Ruby has no concept of casting or blessing the object into a different class, so I suspect that I'll have to take care of that by wrapping the db4o api with some Java to do the casting.

I can confirm that something like the objects I'm storing are ending up in the database file with the venerable strings command. Surprisingly the database binary file contains strings that look exactly like that generated by to_s. Perhaps it stores a stringified version of each object? Or perhaps to_s is proffered by JRuby as a default serialisation method.

Anyway, as I said I haven't invested much time on this (too busy watching the Monaco F1 right now...) and getting it useable may be a bit of work.

One weird thing is that the first time I run code like the above for a particular database file I get an ObjectNotStorableException exception. The db file is created though, and further runs work as described above.

I was browsing in Borders yesterday and couldn't help thumbing through a Mythbusters coffee table book. I found the following great quote:

Whenever an engineer builds something that's over-complicated, that's the mark of an inferior design. Simple is always better.

—Jamie Hyneman, Mythbusters

While you can't really rate Mythbusters for statistical significance, I think Jamie's on the button with this quote.

I love seeing someone first use closures for a real problem. If they're a long time programmer who has never used them before it takes a while, their probability filter tries to stop the neurons from connecting, and then with a spark the simple beauty emerges like a glorious dawn.

Share the joy with Jeremy Stell-Smith.

With apologies to Bob the Builder.

There's a great discussion brewing over at Lars' blog about how to keep in a design frame of mind and not get bogged down in implementation details.

Some good ideas coming out, and some mediocre ones from me as well. Come and join in!

NetBeans 6.0 is shaping up as a very good Ruby development environment, and JRuby is quickly becoming a viable platform.

If you're looking for a different kind of beans with a different kind of camping, you might want to check out Big Jim's Camping Beans :)

I love what Thoughtworks Studios have done in Mingle, making it so easy to deploy by using an embedded java database (derby), JRuby and Ruby on Rails.

I also love the concepts why has used in Camping and I figured the three combined would be great for developing small (possibly embedded) websites.

Here's my somewhat ugly bullet point steps to the sample Camping Blog using NetBeans 6.0, JRuby and derby.

• Install NetBeans 6 - M9 is the latest version at the time of writing (caveat emptor - I already logged a few bugs)
  
  
• Unless you have Java 6, copy derbyclient.jar into {Net Beans}/ruby1/jruby-0.9.8/lib
  
  • If you're a Mac, you will find the {Net Beans} directory in "Show Package Contents"/Resources
    
  • You can get derbyclient.jar from your JDK directory if you did a full Java 5 installation, or you can get it in the full derby distribution from db.apache.org/derby
    
  
  
  
• Install the camping gem (NetBeans Tools menu->Ruby Gems->Install New...)
  
  
• Start the derby server in NetBeans (Tools menu->Java DB->Start)
  
  
• Make a new "Ruby Application" project (be sure to call the main file blog.rb instead of main.rb)
  
  
• Paste why's example blog.rb into your blog.rb
  
  
• after the line that says require 'camping/session' add:
  
  gem 'ActiveRecord-JDBC'
  require 'active_record/connection_adapters/jdbc_adapter'
  
  
• Create a file in your home directory called .campingrc with the following:
  
  database:
  adapter: jdbc
  driver: org.apache.derby.jdbc.ClientDriver
  url: jdbc:derby://localhost:1527/sample
  username: app
  password: app
  
  
• In Project Properties->Run, add a configuration called camping, and add the path to the camping startup script into the ruby options field. I added:
  "/Developer/NetBeans 6.0 M9.app/Contents/Resources/NetBeans 6.0 M9/ruby1/jruby-0.9.8/bin/camping", I haven't been able to simplify it with environment variables unfortunately.
  
  
• Click the Run button (or use the Run menu)
  
  
• Point your browser to http://localhost:3301/

And that's all there is to it! Note that instead of bothering to make a new database, I used one of the sample derby databases, but you can administer the derby databases in the Runtime window/tab of NetBeans.

Sources:

JRuby Tutorial - Going Camping
The curse of camping

NetBeans 6.0 is very cool for a lot of reasons, which I'll blog later.

But the jaw-droppingly cool thing is ... wait for it ... the OS X version is distributed as a shar archive!

If you don't know what a shar archive is, then I you probably weren't a Unix/Internet nerd in the 80s/early 90s (which is probably for the best).

This is a hilarious Saturday Night Live sketch around the whole liquids on planes thing :)

Found via Pensieri di un lunatico minore.

Pango and Cairo bindings for Cincom Smalltalk—now that's a kick ass cross platform application development environment!

Umbrello is a great open source UML modelling tool. Unlike Argo UML it supports (a subset of) UML 2.0.

Being native code based on the KDE libraries it's much faster than Java tools (like Argo UML), but that also makes it less portable. I've never seen a workable installation of Umbrello on Windows. There are some partial attempts using KDE 4, and apparently it's possible with Cygwin.

It's pretty easy to install on Mac OS X, as long as you don't mind a bit of commandline work.

First you need Apple's developer tools and X11.app. Then you need an installation of MacPorts which will do most of the hard work for you. There's no port for Umbrello, but it will install the myriad required libraries. Simply run:

and let it do the hard work.The whole process could take a few hours depending on your Mac model.

Now to install Umbrello itself. Download the latest version from http://uml.sourceforge.net/download.php and follow the "Source Code" instructions on http://uml.sourceforge.net/install.php. You will need to specify the qt and jpeg directories and turn off aRts support. If that makes no sense to you, the instructions boil down to:

Now you have a working Umbrello! Simply start X11.app and run /opt/local/bin/umbrello (you might want to add it to the "Applications" menu of X11.app via the "Customize Menu..." option).

It shouldn't be too long before we can have a native Umbrello.app. Trolltech has released a MacOS native Qt and it's certainly something that the KDE people are looking at.

Update: Umbrello 1.5.7 is now released - update your download accordingly.

Leopard Update: A kind anonymous poster gives this ./configure line for building in Leopard, but I haven't tried it yet:

I played with GOODS a little while aback with Smalltalk. Now I *love* relational databases. There are so many real world business problems that relational databases map very well and provide fast access to. I also rail against simplistic object-relational mapping with other developers semi-regularly!

But if what you want to do is store and retrieve objects, then an OO database is, at least in some cases, the most appropriate tool. We all love garbage collection in ram (well - I do anyway!), why not let it take care of your persisted objects? And who would say no to the database automatically upgrading accessed objects when your class gets new or different instance variables?

While it's not quite with the theme of the intended use patter of GOODS, I'm going to play with Pogo - a Perl API for GOODS. Unfortunately I don't yet have it compiling on Mac OS X but it compiles fine on my Linux box.

One of the problems with mapping a Perl or Ruby class to an OO database is that we can't reliably introspect the instance variables (let alone their type). I'm not quite sure how to handle that yet - whether there should be an API to make promises about your object. Another problem is that variable types aren't fixed. The biggest of this class of problem is that, especially in Perl, it's not uncommon to be able to use an object or a primitive (string etc.) interchangably eg. you might provide an URL object that can be used like a string in string context, but in reality it's a reference to an object - ideally we'd like the whole object to be stored.

Being more Smalltalk like Ruby may be better suited, but unfortunately there's no Ruby bindings for GOODS and I don't have the time or the in-depth understanding of the Ruby object internals to pull that off right now. There are a few references to a (now defunct) source forge project ruby-goods, but sourceforge's CVS archive shows no files.

I'll keep you posted!

I found a site where you can buy small run design products by small design firms (they may prefer 'boutique' or similar ;). I especially like this shoe rack (which I think I saw linked on trend hunter), but there are lots of other great designs as well.

Their product selection is of a similar vein to Top 3 by design (for my Australian readers). Check it out: GENERATE.

Random Good Stuff has been in my must-read rss list for a long time. I mean who doesn't like random good stuff like geeky clock radios, portable BBQs and Donald Trump selling steaks?!

In the 2007 netosphere it doesn't get much more good than Joost and guess who is giving away invites?

The boys at TramTown (Semi and DB) have just notched up 4 years and 3,005 posts on their blog.

Well done! Maybe May should feature a "best of TramTown" series :)

They have kept remarkably true to their initial post:

This is a blog where Semi and DB are going to put pointers to things that interest them and may interest others. We expect a bias towards our town, Melbourne, Australia. We expect to see posts about: theatre, audio, politics, motor racing and any other thing that tickles our respective fancies.

That's what's good about aiming low.

Wouldn't that be fantastic! I first heard of this in an Economist.com Singapore briefing:

A local property tycoon is bargaining to bring Formula One to Singapore. Ong Beng Seng has been in talks with Bernie