Ben Hughes

I promised myself that I wouldn’t be drawn into blogging or tweeting about Twitter, but some of the comments flying around about #SXSW “information overload” highlighting the failure of Twitter to scale have irked me somewhat.

This is a basic problem in any information management system: you need to adapt how informaton is managed based on the volume and other characteristics of the data. And it’s not necessarily something that can be automated, because information categorisation happens to be one of those things that humans are very good at.

If I ran a phpBB forum with a single forum for all posts, and saw a sudden increase in posts that made following the content difficult, I wouldn’t run around screaming about the inability of forum software to scale. I’d just split  the main forum into several sub-forums, based around the main topics.

If I had a single pile of papers on my desk and found it’d grown to a size where I could no longer find things quickly, rather than declaring that piles of paper were dead, I’d split the pile into several categorised smaller piles.

This is nothing new.

The issue seems to be that people have forgotten how the Twitter syntax has developed over time. @Replies (@<username>), retweets (RT<message>) and hashtags (#<topic>) were all conventions started by users, which virally became popular, and then started being supported by the various Twitter  clients. This is a demonstration of the power of twitter: it’s minialism means that people can use it in very different ways, and the community can extend and adapt it without needing to convince a central authority.

So calling for ‘new analytical infrastructure’, to me, is forgetting the essence of Twitter as a community driven medium.

So as a member of the wider Twitter community, what are my ideas for a solution? First you have to look at the problem a little closer. Hashtags are used in two ways: to allow your followers to easily access the wider context of your tweet, and to allow people following the wider context to find your contribution. The latter is a great information mining resource, and this is where the complaints around SXSW are stemming from.

I think the most obvious solution is similar to the examples above: either use multiple hastags to multi-categorise tweets, or use hierarchical hashtags to sub-categorise them. For a large event like SXSW, sub-tagging would allow much better filtering. For example, using #sxsw:social for social events. 

Why would someone be motivated to use up more valuable tweet space with extended tags? For the same reason they use hashtags now: to improve the quality and usefulness of their stream to followers, and to surface their tweets to a wider community. Category overload is as much a problem for contributors who want to be found as it is for consumers looking for data.

If you happen to be following a large number of people who are attending something like SXSW, some enhanced Twitter tools might be nice. For example, being able to filter out tweets with a particular hashtag, or in-built group support and the ability to temporarily ignore updates from a group. But it is possible to handle now, either by using a more powerful Twitter client like TweetDeck, or by actively managing the people you follow. 

Going forward, I think improved filtering is definitely going to be important when it comes to the usefulnesss of hashtagging, whether provied by Twitter or third parties. Right now though it seems silly to be saying that social media tools have failed to scale simply because people have dumped all of their content into a single category. There was a failure of users to anticipate and react to the volume of tweets, but nothing, I think, fundamentally flawed with the infrastructure itself.

A little while ago my main amp - a NAD 3150 integrated - developed a problem in the pre-amp stage which has meant that I’ve been using a PA2V2 portable amp in my main setup. As timing would have it, the PA2V2 never actually made it to work, which is what I bought it for. 

With the immenent arrival of my new headphone amp and DAC, the PA2 will be free to head to work. Before it does I wanted to do some quick tests to see if it actually improves sound quality at all.

The testing setup was pretty simple:

• 60Gb iPod video in a standard Apple dock, playing 320kbs CBR mp3s (lame –preset-insane)
• A/B switch box connected to the iPod’s headphone jack
• PA2V2 connected to the line out of the dock
• A/B switch connected to the PA2V2
• Ultimate Ears SuperFi3 canalphones

After level matching the volume by ear using a 400hz sine wave I started as I usually do: fully sighted subjective listening. It was immediately obvious that a simulated blind test wasn’t going to be needed - there was a very significant, easily detectable difference in the sound.

The first stand-out difference was in the quality and amount of bass. Initially I thought it may have been the amp adding some mid-bass EQ (which it may do) but the sound didn’t become muddy as is usual with in that case. In fact throughout the specturm instrument separation seemed to improve, and the soundstage widened appreciably. 

Best of all, the sound was much more dynamic and lively with the PA2V2. By comparison, the headphone socket sounded flat and dull. Given how little power the SuperFi3s require (I often listen to them on the ipod’s lowest volume setting at work) I was really surprised at the difference even a cheap amp like the PA2V2 made.

Is it enough of a difference for me to carry a LOD and amp for portable use? Nope - it’s just not worth the fuss to me. However it’s definitely worth the price of an iPod dock for work.

I haven’t used Program-line (”P”) mode on an SLR since 3 weeks after buying my *istDS, well over 2 years ago. I quickly found that Aperture priority (Av), and to a lesser extent, Shutter priotiry (Tv), allowed me to control the camera in a much more desirable way. Lately for more contemplative images I’ve been using the Zone system - shooting in Manual with spot metering - and using Av for quicker, kid-chasing type shots. The Pentax-specific TAv mode has been invaluable in variable, low light situations.

But today I was messing around with settings and remembered reading something about the K10D’s P mode being referred to as “hyper-program” mode. Not remembering exactly what that meant, I played around for a bit. And then BAM! Suddenly I was loving shooting in P mode.

Hyper-program mode works just like any other program line mode - it best guesses at the Av, Tv (and optionally ISO) level, changing all variables to get the right exposure You can set biases so it will, for example, go for a short a depth of field as possible. All normal so far.

Until you hit one of the e-dials. Moving the rear e-dial allows you to set the aperture, which stays set until you hit the little green ‘program line’ button. The front e-dial controls the shutter speed in a similar way.

So essentially you can move from P to Av, P to Tv, Av to Tv and Tv to Av by just moving an e-dial. That’s much quicker and easier than moving the mode-selection dial.

To be honest, it’s not that often that I want to switch from Av to Tv. In fact I rarely use Tv at all unless I’m trying to shoot a moving object. But it does happen from time to time. What happens more often is handing the camera to my wife, which usually involves fiddling with a few dials to put it back into ‘green’ mode. Being able to just hit the green dot button to move back to full P mode will make that at least slightly quicker.

Despite the stigma, I think I’ll be shooting in P rather than Av from now on. I just hope noone is looking at my EXIF.

It’s true - I can’t help myself. Once I noticed that the headphone-out on my 900HA was much cleaner than I had expected, I decided I needed to do some testing. Nothing too serious, as I have no plans to actually use an eeepc as a music source, but enough to satisfy my curiosity.

The setup was as usual: both sources connected to my DIY switch box, then into the PA2V2 amp and the MS Pros. I level matched the sources by ear using a sine wave, ending up with the ipod on full volume  and the 900HA 4 steps below full volume (jack set to line out mode). The 900HA was booted into Windows, as the audio seems to be better quality. Both sources were playing WAV files.

The A/B testing was done fairly subjectively, without the usual rigour. Rather than having a certain number of guesses and recording the results, I just listened to a dozen tracks or so, switching back and forth, checking my guess, then re-randomising the input selection.

Suprisingly both were quite similar in sound, and it wasn’t until near the end of the listening session that I identified the different sound signatures. The ipod was noticably smoother, especially in the mids, whereas the  900HA had a more live/harsh sound. By the end of the session I could pick the source around 8/10 times.

Despite being still very impressed with 900HA, the ipod subjectively has a much nicer sound. But it’s impressive that a netbook that cost only slightly more than the ipod is so close sonically. Especially when every full-size laptop I’ve heard has utterly terrible audio.

As an aside, synching the tracks between sources was a bit of a challenge, as both had a slightly different delay from button-press to track start. I eventually got the swing of it, although never got it perfect. It’s much easier to test amps than sources.

As a follow-up to my last post on installing Ubuntu on an eeePC 900HA I thouguht I’d do a quick write-up on eeebuntu, which as you  can tell from the name is a Ubuntu derivative customised for the eeePC. The current version is based on Intrepid Ibex, so you don’t miss out on any of the cool stuff.

The main benefits of eeebuntu over Ubuntu is that the majority of the hardware has support compiled into the kernel. That means no messing around looking for wifi drivers and having most of the function keys work without fiddling.

An added bonus is that Gnome has been minorly tweaked for the smaller resolution screen, it’s got a good-looking theme and they have bundled quite a few  useful apps, like CCSM, Pidgin, msttcorefonts and the OSX-dock-like AWN.

There was still some minor tweaking needed after the install, like changing the fonts to verdana (or one of the other true type options), enabling hibernation (via gconf-editor) and setting up Compiz to my liking.

Unfortunately the 900HA is not yet supported by the “eeebuntu config” utility, however the vast majority of things seem to work. One thing that doesn’t is CPU scaling. This seems to lead to around a 1/4 loss in battery time over XP (which uses the bundled ASUS power management app). For general browsing with the screen brightness turned down a fair way I can get around 3 hours under eeebuntu and at least 4 under XP.

These days I very rarely boot into Windows, unless the extra hour battery life is vital. One thing to keep an eye on is Firefox CPU usage - for certain sites (and with some plugins) CPU usage went up to 100%, which is a sure fire way to kill battery life.

Overall I’m very pleased with eeebuntu - it’s slick, works out of the box, and introduced me to some cool things like AWN. Highly recommneded for anyone running an eeepc.