People who know me, at all, will know I love two things: cheese, and the colour blue. Oddly enough, I don’t like blue cheese, but that must be some kind of cancellation effect.
Until the availability of web-based cheese, I have instead opted for sticking more blue in the site. Unless you’re using one of those newfangled ‘arr-ess-ess readers’, you’ll have noticed by now.
In other news, I’m working on an upgrade to lastgraph3. I’ve finally had a sudden onset of common sense and have modified the downloader to not be ’scalable’ and so not fail, am storing data in pickles rather than a database (it’s way faster and more efficient, since the data is denormalised). There’s also a more interactive interface, including browsing of artist-specific listening histories and weekly artist breakdowns.
Hopefully it’ll be out by the end of the week, but who knows, what with some exams apparently in the way.
Anybody lucky enough to own an OS2008 internet tablet can be smug and happy knowing that OpenTTD is now available for OS2008.
You need the data files, of course, but that’s the same as usual, non-maemo OpenTTD. On the plus side, I actually discovered I really can still edit C++ without breaking it (I tend to spend a reasonably large amount of time using high-level languages, as I value the amount of typing I do).
I’ve put the slides up from my Lightning Talk about LastGraph I did last week for Oxford CompSoc. They’re almost the same as the ones from Barcamp Brighton, but with a few small changes, since I, er, kinda rewrote it since then.
After comments mysteriously stopped working on my old site, I’ve migrated www.aeracode.org to the Aeracode Network™ from my old host. Hopefully things should be nicer now.
Now, it’s not very often that I write long posts on random internet topics, but this is today’s exception, because there’s something that’s really been annoying me lately.
If you look at the Web, it seems to generally work quite well - it’s only partially centralised, and most of that is at the networking layer rather than the actual HTTP layer, where there is no requirement to rely on any central third party. We’ve had the freedom to host and create for several years now, and it’s done wonders for the Web, with what is quite possibly the most diverse thing in the history of humanity being easily accessible and easy to contribute to.
For years now, we’ve been hearing about how we’re going to take the step from simple 2D web pages to the ‘3D web’. First there was the fantastic flop that was VRML; a combination of obscurity, inflexibility and speed meant it was neglected and used by only a select few sites. Since then, we’ve had the massive explosion in home bandwidth and thus MMO games, and perhaps the most interesting thing to emerge from all of this is Second Life.
For those of you who have been living under a rock for a few years now, Second Life is a rough equivalent of what many people predicted the “future internet” would look like. It’s an interactive, open 3D world, with its own scripting language, fully customisable objects, and quite a few other things.
All in all, it sounds quite good. However, it does suffer from what I see as a few fundamental design flaws, the first and foremost being that it is a single, contiguous world. In other words, it emulates the real world (in perhaps a similar vein to Stephenson’s Metaverse, although obviously they differ in other ways). My issue with this is that the real world doesn’t translate so well to the internet environment (and also has several design flaws itself, but that’s a discussion for another time).
One of the many consequences of this (apart from making server-spanning more difficult) is that there has to be some central system (in this case, Linden Labs) which manages everything, thus reducing the one key advantage of the internet: that anyone with a server can simply hook up and host.
Even Linden’s open-sourcing of the client and server won’t help this problem, and so I believe a different solution has to be found, one which is more sustainable, and internet-friendly. I feel the key is to nick those parts of the web that worked so well - there’s no need to change them.
Firstly, get rid of the one solid virtual surface. Do what websites do, and split the ‘virtual universe’ into discrete spaces, each of which can be hosted on its own server. We can use URIs to identify these spaces - hostnames give the server, paths the particular world it hosts, and even anchors for specific points in the world. Obviously we wouldn’t use HTTP, but using identifiers like “3d.aeracode.org/gallery” would both mean we could use existing infrastructure to locate servers (no need for another central registry) as well as play on users’ familiarity with these types of addresses.
Then, steal hyperlinks. Since we already have a set of canonical identifiers, we can link people to other worlds; perhaps have objects you can click on, portals you can work on, or other similar analogous objects (perhaps even ‘load point’ like areas, where you’re transparently moved to a different world).
Of course, to make all this work, we need some way to define worlds and objects within them. This is a matter best left up to discussion; whether to go for an easily-readable XML-ish format, or an efficient binary format (or perhaps gzipped XML), would take a bit of debating. Throw in a scripting language for interactivity as well - we can learn from LSL and JavaScript, both of which are good for different reasons.
This is not as easy as it may first sound, if you’ve not tried these sorts of issues before; unlike (most) websites and VRML, I’m proposing a system that users all interact with at once, within the same instance. Synchronisation issues, security issues, and similar obstacles all present themselves; in my previous attempt at a quick-and-dirty MMO engine I came across more obstacles than I can describe here. However, the game industry has been tacking these for years, and so they’re not insurmountable.
We also have the issue of identity. The OpenID movement is a great example of what needs to be done here, although I don’t feel it could just be used directly, as it’s very much biased towards website services. Still, follow the same method of OpenID (and of my previous discussion) by using URIs as identifiers. Identity servers should also act as a kind of data and preferences store for the users, much like IMAP with email, so they can set their avatar/email/’bookmarks’/etc. only once and have it appear on whichever client they use.
Finally, there’s the issue of 2D versus 3D. Everyone thinks that these sorts of environments should be 3D, and while that’s certainly true for the vast majority of devices, there’s something to be said about having a 2D world protocol as well, with the growing popularity of low-power devices like mobile phones, UMPCs, Internet Tablets, and general cheap, small hardware.
There are many more issues that can be tackled, and it’s quite a massive thing. I have bits of code scattered over my code folders with ideas and half-hearted implementation attempts at all this (of course), and it’s somewhat unlikely all this will ever come into fruition - but, hey, it’s a nice idea. Perhaps one day I can do webworld designing alongside, or in lieu of, website designing…
LastGraph is still undergoing rewriting after the beta showed up some lovely bugs (including inverted labels and labels everywhere they shouldn’t be). I’m working on it over the weekend, after my week was overtaken by the sheer mass of activity involved with university and Fresher’s Week (I find it strangely ironic that it’s actually more timeconsuming this year, when I’m not a fresher).
I’ll give a progress report soon; hopefully I can clear up the fetching errors (Last.fm have also kindly agreed to let me access the API more often, so fetching speed should increase), and make uploading to S3 faster (aegis’ upload speed is somewhat limited, so it will be done directly from the render nodes themselves - this does somewhat scupper my plans for distributing the rendernode client, though, as they need my S3 secret key to do this, unless they can upload elsewhere and pass back an appropriate URL).
Update: A beta version is now up at aegis.aeracode.org - it has BUGS, like bad label placement (and inverted labels on some Acrobat installs), as well as rendering VERY SLOWLY while I try new optimisations. Bear with me, please.
Yes indeed, the new version of LastGraph has been mostly written over the past couple of days and should hopefully be ready to go live tomorrow.
Both the frontend and backend have been completely rewritten, and should hopefully be more stable and efficient. The label-fitting code has been drastically sped up (this was one of the bottlenecks), as well.
I’ve moved to Cairo for rendering; it has the advantage of natively outputting both SVG and PDF. Oddly enough, PDF sizes have plummeted now they’re being written fresh rather than converted; a year and a bit of my history is under 200kb, which is a pleasant surprise indeed.
The graphical style has changed a little, mostly due to the fact that I’ve rewritten all the rendering code and so some things aren’t quite how they used to be.
It is also quite likely that there will be some bugs in the first few days; hopefully this can get sorted out.
I’ve moved to using Amazon S3 for storage, so I’ll have no more hard disk space worries, at least.
As a small ‘gift’ to tide things over, here’s a sneak preview of the new graph output, as a PDF.
My headphones broke last Tuesday, and so I naturally went to buy some more. However, my eye caught this post, and since I actually had some spare cash just begging to be used, I decided to go ahead and buy some Jabra BT620s headphones.
They arrived the following day, and so I followed the instructions and got them working under Maemo using Kagu as the blog posting describes. Performance is only OK; there is skipping once or twice, and if you even tempt Kagu’s interface into moving expect lots of blank sound.
This is, I imagine, mostly because Kagu has an effects-heavy PyGame interface. It’s fine when all the music playing is offloaded to the DSP chip, but a2dp requires a daemon so sit and transcode all the data in realtime.
I may attempt to write a more CPU-friendly GUI to the Kagu libraries, like I was planning to convert Spindle into anyway; we’ll see. For the moment, though, I’m quite happy with the playback overall.
The slides from my talk at Barcamp Brighton 2007 are at aeracode.org/files/lastgraph.pdf; for some reason slide 4’s PDF weighed in at 9MB, so I apologise for the large download size (I am suspicious that Inkscape rendered the raster image in that slide to a pdf pixel-by-pixel).
Inkview seemed to hold up quite well, and was plenty enough for this simple talk. I may start promoting it more now.
As good as OpenOffice is getting these days, I have a personal issue with Impress (and, indeed, presentation software in general, such as the ever-popular Powerpoint).
My slideshows are very often not presentable in a bullet-point format. Indeed, it seems slide after slide of bullet points can get a little tiresome. Thus, for my talk at Barcamp Brighton this weekend, I felt much more like going with Inkscape. Unfortunately, it’s not too easy to load inkscape files as a slideshow… or so I thought.
Enter Inkview. Appearing in the latest versions of inkscape, you pass it a list of SVG files on the command line, and it opens a window with some basic presentation controls. It’s pretty basic, but it’s just what I was after. There’s no real way of sharing a common ‘master’ background among SVG files apart from the traditional copy and paste (well, as far as I can see), and if you want to keep your sanity you’ll want to make the filenames correct so you can just call “inkscape slides/*.svg”, but it works.
I much prefer it to Impress. This is probably because I don’t really do presentations too often, but… that’s not really too important. If you feel like a bit more graphic flexibility next time you make a presentation, you should probably at least give it a passing thought.
