I haven't even being doing much tech stuff over the last two or three weeks...

Testing

Well, that's entirely false. I have a new job. A fulltime programming job. In Python. I only just escaped signing a "we own your brain" clause which would have meant that I would have nothing to say because it would be owned. But it was modified and I continue to own my brain during off-hours. And occasionally I will continue to use them for tech stuff.

Tech-wise, the brains's been busy with thoughts about test driven development, which I've never really done before. But I've spent large chunks of the last fortnight updating tests to work with modified code. I'm beginning to wonder whether optimally some other programmer should always write the tests for your code. But this may be unsellable: it takes long enough to write tests for your own code.

Online appeals — a review

Andrew and I gave money to the tsunami appeals: Andrew to CARE Australia and me to the Australian Red Cross. (As a side-note, why I am so susceptible to guilt trips? I keep seeing "only ten thousand people have donated through Amazon!" guilt trips around and falling for them. Note to self: failure to donate through Amazon is not any meaningful kind of failure.)

Within one second of my provision of my CC details to those nice folks at the Red Cross an 'official receipt' email was on its way to me thanking me for my donation and telling me how the Red Cross is intending to use the funds. Nearly a week after donating to CARE, Andrew's still waiting for them to actually process the donation and bill his card. (I rather suspect their online billing handlers too the Christmas-New Year week off...)

Ambition

It's time I had some. I'm close to declaring 2005 to be the year of ambition, much as 2004 was an explicit break from it.

Fixing the Planet <h1> <h2> evil

I notice that Andrew's recent entry has triggered all kinds of evil on Planet SLUG and to a lesser extent on Planet Twisted. The nature of the evil is two-fold:

1. RSS feeds contain an unrestricted subset of HTML tags, therefore there is a collision between tags the maintainer of the planet wants to use in their own content that goes around the blog entries, and tags the entry authors use.
2. HTML doesn't really do sectioning. At one point I heard XHTML2 was going to have section tags with h tagged heading embedded within them. This is a better idea than having the author of pieces of content needing to know what level of heading tag to use to fit into sites designed by someone else.

The solution I'm thinking of using on planets I maintain will apply special CSS properites to headings inside entries to make them less prominent:

div.blogbody h1, div.blogbody h2 {
	font-size: large;
	text-decoration: none;
}

I should look up how to reduce their prominence to screen readers too. Alas, the fundamental semantic nastiness of having the following in the HTML will remain:

<h2>Blog Entry title generated by Planet software</h2>
<h2>Blog entry <em>sub-title</em> coded as h2 by content author</h2>

Bad SLUG, no slug biscuit

I think Planet SLUG is still sending the character encoding as iso8859-1 while they're actually sending utf8 content. (eg the character 'ó' in Andrew's entry)

Python in Sydney

Alan Green organised a Sydney Python Meetup last week. I'm pleased someone stepped into this slot: I was organising Python meetings a few years ago, ran out of energy during honours and was never really inspired to start it back up because meeting attendance never got above five or so. I hope Alan has better luck: he got fourteen for the first meeting.

Alan has given a full account of the talks, so I'll just note one thing: the presence of Tim Churches, an epidemiologist from the New South Wales Department of Health, who was behind the open source release of NetEpi. If any of the state government's employees would like to follow suit, it took Tim 18 months and 37 memos to get the release past the government. Beat that!

But the real reason I was particularly interested to meet Tim was that he's the first example of a phenomenon Anthony Baxter tells me he sees all the time: the professional who just needs a dab of programming and chooses Python because it's ideal for people who aren't specialist programmers. Apparently it's increasingly popular among scientists in particular.

Ubuntu among the Python programmers

I wanted a pressed copy of Ubuntu rather than having to download one, so I went to get one for free. They helpfully informed me that I might as well get a bunch since the cost of shipping massively outweighs the cost of CD production. So I got 10 i386 CDs and 5 PowerPC CDs. I was at a loss for what to do with them — I wore out my parents' tolerance for installing stuff on their machine when I was 12, and most of my non-geek friends are now canny enough to do the "only if you install it and fix it for me whenever it stuffs up!" trick with proposed Linux installs — so Andrew took them along to the Python meetup and gave them away.

Alan was most impressed that the LiveCD worked on his laptop — alas that with the 4.10 release (Warty) of Ubuntu that's apparently absolutely meaningless when answering the question "will the installer work?"

Tim was pretty dismissive of Ubuntu though on the "I tried it and it doesn't even have a compiler!" principle. This isn't actually true — gcc isn't installed by default, but it's supported and I believe it's on the CD — but since it cost Ubuntu a user I'll have a closer look at why I think he thought this.

I wasn't really party to it (I joined the pre-release test team after they decided not to include a compiler in the base install) but as I understand it, the idea was most users won't want a compiler in their install because most users aren't programmers and the idea behind distros like Ubuntu is that you shouldn't need to compile software for it. Further, and this seems to be key, people who want the compiler will know where to get it. Or I assume that was their premise, perhaps with the addition of "developers will read the documentation." (I'm not sure though, "where is gcc?" hasn't made the FAQs.)

Anyway, it turns out that this isn't the case. There are developers who, if the compiler is not installed by default, do not think to try and install it because they have no inkling that it might be there. My suspicions about the reasons:

1. they aren't former Debian users who are used to their distro having every conceivable piece of Free Software available in a centralised archive: what the installer offers is what there is; and
2. they are not expert Linux users, by which I mean that they don't have server sysadmin or security skills. It's quite hard, in fact probably pointless at this time, to program on Linux without knowing a shell reasonably well, but it is possible to program on it without knowing much about how software gets installed on it. You put a CD in, perhaps select "development workstation" and it installs stuff. When you want newer software, you do a new install. When you want different software you might look at a different distro.

I might be reading too much into this, and in any case developers, whether power users, sysadmins or none of the above, aren't the Ubuntu market, but assuming that anyone who programs using Linux has Linux expertise above that of a desktop user is probably wrong.

Clarification re Ubuntu and compilers

Some questions imply that I wasn't terribly clear about this, so now I will be clear: Ubuntu has supported binary gcc and g++ packages. You can install them via apt-get/aptitude/synaptic. (By the way, on Debian systems I use aptitude now because it especially marks "packages that were only installed because I installed another package that depends on it" and automatically removes them when nothing depends on them any more. Even better than deborphan.) It's probably easier to grab the build-essential package though, which drags in those and make and a few other things as dependencies.

The 'problem' with the compiler is not that it is not packaged and supported, it's just that if you whack a CD in your machine, choose the default install, and then log into your brand new Ubuntu machine, you will find that the compiler is not yet installed.

This is not something I personally consider a problem, possibly because I'm a Debian user and also because while I have tried the RPM distros I came to them rather late anyway (RH8) when apt-get like tools (yum, apt for rpm) were not far away. I'm very used to the idea that software that's not installed yet is just sitting in an archive or on the CD somewhere for me waiting for me to use a nifty tool to download and install the package. (As for satisfying dependencies of packages, my automatic reaction on any distro whatsoever now is "what's the apt-get equivalent on this one?" It pretty much always has an answer now too.)

It appears though that some people don't think this way. They think "either it's on the computer at the end of the install or it's a major disaster involving downloading the packages myself and resolving the dependencies by hand or it's a super-major disaster involving downloading the gcc source and trying to bootstrap a gcc compile." They don't imagine this fourth option where it's not on the computer yet, but hey, it's just there on the CD waiting for me to type "aptitude install gcc". (Actually, there are people who do realise this and just hate the idea, but the entry wasn't about them.)

And these users are not a good fit for this particular design decision of Ubuntu's. Which is a pity actually, because I thought it was rather a good one myself.

Google adds a 'no-follow' tag

I haven't seen this mentioned in many places at all, perhaps everyone follows the Google blog?

Google has added an unofficial rel attribute value to the a tag that their indexer will know how to understand: rel="nofollow". Using this tag as part of a link will say to Google "I am linking to this page, but that does not mean that I think its rank in your search engine should be increased."

Why are they doing this? Well, if you haven't heard about comment spam this might not make sense to you, but many popular weblogs and wikis are continually hit with people leaving comments like "get your discount viagra here" with a link. They sometimes get hit so much their bandwidth limit is reached or their server crashes under the load.

People occasionally ask "Does this really work? Sure it appears on the page, but it seems even less effective than email spam. People from my rightist libertarian feminist discussion board have Viagra suppliers already! They will ignore these dumb comments you leave." And that's all very well, until one day you get hit with 15000 spam comments in a single day and have to turn comments off.

The answer to "does it work?" is suspected to be: it works, but not like you think. While they probably would think it's a bonus if your weblog readers clicked, you know who they really want to 'click'? Google. Why? Because Google counts the number of links to a website as a measure of how 'good' that website is as a search result. If they can add their link to enough pages (including yours, through your comment form), then zooooom up the search results they go.

Anyway, this tag is Google's attempt at an answer. By marking URLs provided by commenters as no-follow, you can say "Dear Google, I don't trust this link, so neither should you. For popularity purposes, this link never happened, OK?" Ideally, it will remove the spammer's motivation for posting on weblogs. (The problem of wikis is much harder, since all the content is user created, it's not really possible to mark out bits of it as 'Google good' and 'Google bad'.)

Ubuntu's Hoary Hedgehog release

I upgraded from the stable Ubuntu to the development version (due to be released in April as, I believe, 5.04) on the weekend. So far I haven't actually noticed very many dramatic changes. Booting is marginally faster. There are now suspend scripts that don't work (well, for me, there's positive feedback about them in general). The default fonts look different. Otherwise everything seems pretty much the same.

One thing I was reminded of, particularly when upgrading my workstation which has such a slow disk that the UML server that runs puzzling.org can install packages faster, is that desktop Ubuntu systems have all the Python development libraries installed. It's nice that they're all supported, but I've developed using Python as a primary language for four years or so and I've still used only five of those libraries (more once Twisted splits into packages...). I'd be happy to install them as needed, just like I would gcc.

Mutual obligation

Alignments of the stars this week have me going to a meeting at SLUG on Friday to try and organise a community program that tries to get local volunteers together to work on Free Software. Why add this layer between local volunteers and Free Software, given that most projects have open development communities?

Well, there are a couple of reasons. The first is that to receive unemployment benefits in Australia you need to fulfil a 'mutual obligation' requirement in which you seek work, train or give back to the community. Setting up a local community group/non-profit will allow people on the Newstart allowance to work on Free Software (and docs and bug triage and mailing lists and...) as part of their mutual obligation. The second is that there is a group of somewhat nebulous size who would 'love to help out' but aren't for whatever reason suited to jumping into new communities and offering their work up. Having a group set up to specifically push them into projects may help here: it will be interesting to see.

This is something that's been in the background of my thoughts for years now: working with people who aren't used to the standard model where you just start doing stuff; or the other standard model where you hang out on IRC for a few months until people know you're not an idiot, and then you start doing stuff. This third group is the set who 'need an invitation', if you like.