The UK courts project is floundering. A few years ago ICL won the contract to modernise the UK Courts technology systems, by writing an entire new system for document sharing etc. This was heralded as a great move for the entire country, as many cases are delayed because of paperwork errors, leading to an increase in the numbers of prisoners having to be held on remand, and increasing the risk of criminals re-offending before cases are heard.
As the only company in the running, ICL signed an £183m contract. 2 years later they decided that it would take 12 years, instead of 8, and should actually cost £319m. They also cunningly wrote the contract so that somewhere in the region of half of this be triggerable for setting up the infrastructure for this (installing Microsoft Office on some new PCs).
But, in a recent memo, the Lord Chancellor’s Department has said, “Despite the best efforts of all those involved we have been unable to reach an agreement with Fujitsu on a proposition for Libra which represents value for money and which we can afford.”
If this does all collapse, then UK taxpayers will have paid ICL/Fujitsu over one hundred million pounds to set up some computers and install Microsoft Office on them – in many cases meaning that courts workers have to have two computers on their desk – one to access the court records on the old system, and one to do their email and word processing etc!
Time to write to my MP, methinks.
Interesting presentation from Tools99 on improving code quality through Human Engineering.
The examples are a bit weak, and I suspect that the “conclusions” weren’t so much conclusions, but the starting point. But they’re good:
- Humans make mistakes (especially programmers)
- By “Human Engineering” the language various Sins of Omission, Confusion, and Sloth can be caught
- Human Engineering takes conscious effort:
- Use Redundancy, Compartmentalization, Completeness Checks,
Explicitness
- Anecdotal evidence that good human engineering can significantly reduce test and integration phase
- Much more experimentation and research is needed
[via Ziggy]
In “Agile Software Development Ecosystems” there’s an interview with Kent Beck, where he rants about requirements creep. However, unlike most project managers etc who similarly rant about requirements creep, Kent says: “We don’t want requirements creep; we want requirements to get up and dance. The problem is that they are just creeping. New requirements are caused by evolving businesses and you want as much of that as possible.”
Dave Winer has created a Google Outliner Browser: It’s a different way to crawl through The Mind of Google
For those without outliners, I’ve knocked up a simple web-based version at http://www.kasei.com/google/browse.
One of the really neat features about Wikis is the back linking – click on the title of any page, and see what other pages link to it. I was talking to Ken MacLeod about this a few days ago, and how we could apply it to weblog entries (after the conversation flow problems I talked about a few days ago).
Now I see that Joel has added automatic linkbacks, beating me to it! Nice.
Ebay’s latest financial reports reveal that:
- over one and half million new items get listed every day
- these generate almost $2.5m per day revenue
- on top of this they earn about another $10m per month from third parties (advertising and other services)
- they spent almost $25m in the quarter on the development of the software behind site and seller tools.
I’ve never really understood how companies like this spend quite so much money on “product development”. The last time I added it up (about a year ago), Amazon had spent over $1bn on this (although they include content licensing costs in theirs). Ebay’s run to date is coming up on about $200m. From what I can see this doesn’t include costs for hardware or hosting etc (other than for equipment used for development) – the majority of it is staff and contractor costs.
$25m/quarter pays for the equivalent of 1,000 developers at $100k/year each. Take off even a sizeable chunk for their equipment, and make some of them managers and support staff, and you still have a ludicrously large team. As I don’t believe they have anywhere near that many developers I can only assume that a huge chunk of this is going on software licenses.
Amazon announce last year that moving to Linux saved them around $17m in technology expenses in the quarter. At BlackStar we managed to keep lots of our costs very low by building everything around a LAMP platform. I think some other people need to learn these lessons…
In Q4 2001, Amazon managed to save $22m in costs by tweaking its fulfillment operations. As well as the obvious fine tuning of its predictive software, it seems that one of the most effective changes was to rearrange the layout of their warehouses to make it quicker to pick goods that often get ordered together. I’m not sure what percentage of that the $22m that accounted for, but even if it was 10%, that one piece of software will save them somewhere in the region of $10m in a year.
Nice return for a fairly simple idea…
I discovered a wonderful Template Toolkit plugin yesterday: XML::Style.
The basic idea, according to the docs, is that you can apply various attributes to your HTML. The example given is of transforming an HTML table:
[% USE xmlstyle
table = {
attributes = {
border = 0
cellpadding = 4
cellspacing = 1
}
}
%]
[% FILTER xmlstyle %]
<table>
<tr>
<td>Foo</td> <td>Bar</td> <td>Baz</td>
</tr>
</table>
[% END %]This didn’t sit quite right with me though, as that seemed to be something you should be doing in CSS. But as I read through the docs I discovered you can also change tags. Again though the docs gave a bad example:
[% FILTER xmlstyle
th = {
element = 'td'
attributes = { bgcolor='red' }
}
%]
<tr>
<th>Heading</th>
</tr>
<tr>
<td>Value</td>
</tr>
[% END %]Having been playing with XSLT recently, though, a lightbulb went off. The real power of this plugin is more to be able to do things like:
[% USE xmlstyle
video = {
pre_start = '<html><head><title="Video Info"></head><body>'
element = 'table'
attributes = { class='videoTable' },
post_end = '</body></html>'
}
title = {
pre_start = '<tr><td>Title:</td>'
element = 'td'
attributes = { class='videoTitle' }
post_end = '</tr>'
}
price = {
pre_start = '<tr><td>Price:</td>'
element = 'td'
attributes = { class='videoPrice' }
post_end = '</tr>'
}
%]And then, given some XML such as:
<video>
<title>La Double Vie De Veronique</title>
<price>10.99</price>
</video>We end up with:
<html><head><title="Video Info"></head><body><table class="videoTable">
<tr><td>Title:</td><td class="videoTitle">La Double Vie De Veronique</td></tr>
<tr><td>Price:</td><td class="videoPrice">10.99</td></tr>
</table></body></html>This could be used as a first step towards “true” XSLT if you’re already using TT. If the only reason you’re moving towards XSLT is because a PHB says to, it might even be enough to convince them that you’ve done so :)
Scott Johnson is not a moron. I must be. I had exactly the same problem as him, but never solved it. As I said to Dave Winer at the time: “I’m obviously missing something really obvious here…”
But, I’ve changed my mind. It’s not at all obvious. “Choose the Open command from the File menu in the Radio application”. I never dreamt that “the Radio application” has to be launched through right-clicking in your status tray. I had noticed the Radio icon, but double clicking it merely brought a webbrowser to the fore. Like Scott I assumed that the Radio application was browser based. And I just could not find that menu. Unlike Scott I gave up.
I tried subscribing to the help list first, but it’s terrible. Every mail comes from the same address, and the first 30 character of the subject are taken up by “[radio-userland] New Message: “. In my mail client that means I only get about 5-10 characters of real subject line context. And each message is really just to let me know that “a new message was posted” to the help board, so there’s no “In-Reply-To” headers, or anything that will let my mail reader thread the messages.
Basically the list is unreadable.
So, then I gave up.
I suspect most users won’t even get as far as I did. There is indeed a fundamental conceptual disconnect with how Radio portrays itself.
I like outliners, and wanted to find the one in Radio – but not enough to work through this.
Scott Andrew has a nice drag and drop window demonstration. Unfortunately there’s a nasty bug where Mozilla can’t drop the box, but the concept is good. There’s lots of little loose ends that he’s tossing out for other people to fill in, which is good too. This is already very neat, but it could be exceptional when it gets fleshed out a bit more.
