Thursday 23 January 2014

Statelessness and Chromebooks

I’ve periodically ranted on about how Chromebooks make an ideal machine to take travelling and so on because they are pretty near stateless - all your data is elsewhere, so if you lose it all you lose is your local device - annoying to be sure, but equally as all your data in Googleland, it’s not as if you can’t get access via some other device.

Up to now all this ranting had been pretty theoretical. Then a week last Sunday, my Chromebook decided to shred it’s internal disk. It made an audible click and stopped. Cycling the power caused it to try and boot, get as far as the splash screen, and crash again with an audible click.

Forcing it to go back to factory default didn’t work. It didn’t stay up long enough for the magic key sequence, neither did pulling the battery out and waiting 10 or so minutes before replacing it.

It was dead. Deceased. That moment when a useful device turns into a useless lump of plastic and silicon.

Fortunately it was under warranty, so I phoned up Acer’s technical support line to lodge a warranty claim. They of course tried the standard twenty questions on me, which descended into comedy after have you switched it on and off? and is it charged up?, as the helpdesk person started on the windows laptop not the chromebook Q&A.

That resolved, they agreed it needed a return, and duly emailed me a prepaid label. I’ve got to say they were pretty efficient both at fixing it and keeping me informed.

When I got it back yesterday it had been reset to the factory default out of the box configuration. I, of course, with the arrogance of all IT professionals, had not bothered to make up a recovery USB of my personalisations.

First of all it appeared not to boot, but that was just me being impatient, it was doing a self install/self config. Once that was done getting back on teh road was a case of simply reentering the network configuration information, my google account details, changing the wallpaper, letting Chrome reinstall the evernote, pocket and buffer plugins that I use and we were done.

Ten minutes from boot to working. In the meantime the OS had updated itself so another 90 seconds for a reboot and we were done.

The thing that took longest was deliberating over my wallpaper.

Now my Chromebook shouldn’t have shredded its disk like that, but I’ve been around computers long enough to know that it’s a case of when not if.

As the Chromebook was effectively stateless I didn’t lose any data, all I lost was some network configuration and personalisation data, neither of which was a big deal to put back.

A long time ago, when God was still in nappies, I designed a network solution using PCNFS. All the applications and data lived on servers, and the PC’s loaded everything from the network. The local disk was simply a cache (actually it was a little more complex - users could have private applications locally and the login script checked for them and their config data and added them to the desktop at startup. There was also a local install of the OS so that if the machine was off the network it could boot and use local applications)

As a solution this was extremely powerful. All the data was stored centrally and backed up. Application upgrades could be performed in a single central location. If a machine’s disk died, all that was required was a disk swap. Support and reconfiguration was vastly simplified reducing technician costs.

Changes in the dominant operating system killed this design - and we then started to have to leap through hoops imaging machines and the like. What I like about the Chromebook solution is it’s elegance and simplicity - your data and applications live elsewhere that’s very reliable. The thing in front of you is simply an access device - a portable thin client …

Written with StackEdit.

Wednesday 22 January 2014

Altmetrics and impact (again)

I’ve been thinking further about altmetrics and impact, and how they relate to scholarly publication.

As we know scholarly publication is a numbers game. The more a publication of yours gets cited, the better a researcher you are deemed to be. Get consistently cited and you get tenure.

Citation indices provide the numerical evidence underlying this, and as we’ve discussed before, the bibliometrics people try and provide some weighting to account for some journals being seen to better than others.

This is in a sense simply the formalisation of tacit knowledge. We know for example that a paper in Psychophysiology can be said to have a greater impact among psychophysiological researchers than one in a less important journal, in part due to the reputation of the editorial board, and their desire to retain their reputations by not letting any gonzo research into their journal.

In fact, in a relatively small group such as psychophysiology researchers there is a group of eminent researchers who function as guardians of integrity due to their positions on editorial boards and learned societies.

In a larger field, such as organic chemistry, this is not quite the case, but there is enough of a network in place to provide reputation checking.

Of course, in trying to quantify the relative worth of journals the biblometrics people have created a pressure for researchers to publish in more eminent journals to get a better score. This forces publication towards particular journals, and may inhibit alternative publication such as on arXiv.org.

Now, let’s take a look at altmetrics. Altmetrics really attempts to provides some sort of numerical basis to guage impact from non traditional publication, commonly thought of as blogs twitter and the like.

There’s a little problem here. If a research group blogs about their latest results that’s not unlike announcing results at a conference. If someone outside the group blogs about the announcement its really a measure of public engagement, not scholarly impact. Ditto for twitter.

It’s telling us that the public are engaged by the research, but it doesn’t measure the impact of the research within the discipline.

As an example, we know that for various complex reasons (education, culture, direct relevance to personal circumstances, degree of specialist knowledge required) the general public tend to be more interested in content relating to history and archaeology than plant ecology, even though the latter field might tell them things about something of public interest, such as climate change, and consequently any measure of engagement will reflect this.

This could be considered immediacy of impact.

Let’s also be clear that I am not arguing that altmetrics is flawed, rather that it tells us something different about impact, perhaps reflecting the public perception of worth. It also perhaps suggests that researchers need to consider how better to communicate the relevance of their research, even though it may be very complex …

Written with StackEdit.

Monday 20 January 2014

Windows and me

It surprises a lot of people I meet that used to be a Windows wrangler. For some reason they seem to expect some bearded sandal wearing linux geek or else some wide eyed Mac evangelist.

Well despite increasingly feeling unhappy about windows start up times and upgrade mechanism, that’s not the case. In my time I’ve written registry poking code and code to poke active directory as well as supported the Windows desktop.

While I used a Mac at home - well until the late nineties anyway - at work it was Windows all the way. I even argued for abandoning formal support for Apple when it looked like they were going to go down the toilet in the nineties.

As a manager of a managed (however loosely) computing environment I always found Apple impossible to talk to, to engage with, while Microsoft, well you could talk to Microsoft. You may not have liked some of their behaviour, especially their licensing models, but you could talk to them, and their behaviour was always rational, aimed at obtaining the maximum revenue for their shareholders.

And remember, it’s only recently that Microsoft have had serious competition. In fact I’d argue that they still do not have massive competition. What has in fact happened is that a combination of the GFC and the consequent contraction of business spending, plus the advent of the iPad has simply taken a large chunk out of their revenue.

Basically, the GFC has meant that businesses have kept computers going for longer and lengthened their upgrade cycle. The disaster that was Vista gave them a reason to stick with XP, and Windows 7, while considerably better in user appeal than Vista didn’t quite grab back everything that went to XP.

At the same time the advent of the iPad and other tablet solutions meant that a large part of the consumer market decided they didn’t need a PC. After all an iPad letes them do their online banking, tweet, play with facebook, surf the web and do a little bit of email perfectly well. And you can carry it about, it’s light, you can use it anywhere you have a connection.

And while OS X has also taken a chunk out of Windows, purely because Apple machines feel nicer, I have this suspicion that Apple have cannabalised some of their own market with the iPad.

People on the whole want a low cost lightweight computing solution - the netbook wave was the first part of it, which Microsoft hobbled by the memory and restrictions in Windows 7 Home Basic - on a different day a different decision might have given them a platform to fight back against the iPad with. Certainly the increasing popularity of the ChromeOS platform suggests that people still have an appetite for a low cost computing device with a keyboard, not to mention all these bluetooth iPad keyboards out there.

And Linux - well linux has stolen the server space from Microsoft, and from the high end server makers such as Sun by being able to give good performance on cheap (and or virtual systems) and having little or no licensing costs - simpler to administer for one thing.

But it’s never conquered the desktop. It’s lacked a company that will seriously push and support a linux based solution, and so it remains a mess of different window managers and something that only the bearded geeks among us use. Which is a pity, because it’s perfectly usable as a desktop alternative.

After I gave up on Macs at home in the nineties I used windows, plus a bit of linux both at home or work. I did buy an iMac a few years ago but that became J’s desktop in the way that things do and I ended up using a Xubuntu based machine as my home machine for a few years, until I changed back to Windows. And I changed back becuase the hardware was no longer supported, not because I was unhappy with Xubuntu.

So, I’m not a Windows hater. But it’s definitely lost its edge. And the range of the iPad and web based computing reduces the number of tasks that you simply must have Windows (or Office) for month by month.

Basically, Microsoft now faces competition in a way it hasn’t for around twenty years - and what we’re seeing is the behemoth gradually turning round …

Written with StackEdit.

Saturday 18 January 2014

Ah, windows upgrades

I am increasingly over Windows.

This morning my Windows 7 laptop rebooted four times while it installed upgrades and fiddled about with its registry. When it thought it was done it wasn’t, it had managed to lose the desktop and required a further reboot.

If you use OS X or Ubuntu, you don’t have to do this. The odd reboot and pause but that’s about it, plus some degree of user control over when and where you upgrade.

With ChromeOS of course it’s simply a reboot.

And, of course, as regular users of non-Windows systems know, they boot considerably quicker than Windows …

Monday 13 January 2014

And where would you like your mail delivered sir ?

I’ve previously written about the decline of the postal service

It so happened that I needed to send a package today, and my receipt had the slogan Help us build an even better Australia Post in your community and a URL to go to to complete the survey.

So I did. What was interesting is that it was obviously designed to market test a couple of options:

  • 3 day a week delivery - the New Zealand post model
  • Collect you mail from your local post office - the Canada Post model

Nothing wrong with either option - Australia Post is subject to the same costs as overseas postal services and should be testing how acceptable either would be.

However, what was wrong was the concept that everyone wants to go to their local post office.

It so happens that our designated local post office is in the opposite direction to where we work, and its opening hours don’t mesh nicely with the working day once commuting time has been factored in. That’s why we pay to have a post box at a different location - one we can get to during the working day to pick up packages and so on.

If we are all going to have to collect our mail from somewhere it should be from a preferred location - there’s a number of locations that would work for us, or even better, given that we are already paying for a post box offer us a permanent redirect as part of the package …

Written with StackEdit.

Friday 10 January 2014

using Libre office on roll.app on a chrome book

After my 3g router fun, I finally got round to to testing Libre Office via Roll.app on my Chromebook. The test was pretty cursory
  • open an existing document from Dropbox
  • make some changes
  • save them
  • create a new document
  • write some text
  • save it to Dropbox
To carry out the test I connected using my 3g network connection. I found that in use
  • Libre Office slow to initialise - the processing the letter A problem
  • fine for opening existing documents from dropbox
  • creating new documents fine but typing without naming doc first did not work - this could be due to lack of patience on my part
    • this differs from behaviour using chrome on a mac
    • could be artefact of using slower network connection
I fed the resulting newly created document through Apache Tika to look at the technical metadata in the document and everything looked fairly normal - essentially it’s a recent version of Libre office running on Linux.

apache tika test results

Character Count: 55
Content-Length: 8198
Content-Type: application/vnd.oasis.opendocument.text
Creation-Date: 2014-01-09T21:08:51.491268425
Edit-Time: P0D
Image-Count: 0
Last-Modified: 2014-01-09T21:10:25.257548178
Last-Save-Date: 2014-01-09T21:10:25.257548178
Object-Count: 0
Page-Count: 1
Paragraph-Count: 1
Table-Count: 0
Word-Count: 11
date: 2014-01-09T21:08:51.491268425
dcterms:created: 2014-01-09T21:08:51.491268425
dcterms:modified: 2014-01-09T21:10:25.257548178
editing-cycles: 1
generator: LibreOffice/4.1.3.2$Linux_X86_64 LibreOffice_project/70feb7d99726f064edab4605a8ab840c50ec57a
meta:character-count: 55
meta:creation-date: 2014-01-09T21:08:51.491268425
meta:image-count: 0
meta:object-count: 0
meta:page-count: 1
meta:paragraph-count: 1
meta:save-date: 2014-01-09T21:10:25.257548178
meta:table-count: 0
meta:word-count: 11
modified: 2014-01-09T21:10:25.257548178
nbCharacter: 55
nbImg: 0
nbObject: 0
nbPage: 1
nbPara: 1
nbTab: 0
nbWord: 11
resourceName: 00_apache_tika_test.odt
xmpTPg:NPages: 1


While I was at it I also ran the same test with open office writer - behaviour was broadly similar
  • slow startup/initialisation
  • no need to save document before entering text
Some other testing on my 3g connection seems to suggest that there is sometimes a bit of latency starting up web based applications, certainly I don’t see the same problems when starting up either on Chrome on a Mac or Chromium on Linux on a high speed connection at work.

and the processing the letter A problem?

In the eighties I was employed by the University of York as an analyst/programmer officially working some of the time on time sharing Vax based systems. In fact I actually spent most of my time messing about with these new fangled desktop pc’s, text processing, and data capture and migration - no change there then.

However, my alleged status as a Vax programmer meant that I did get to go to various Digital Equipment Company conferences and at a Decus conference at I think Warwick, a guy called Bill Hancock gave a hilarious presentation on the processing the letter A problem.

The scenario was a Vax in a nuclear power plant that was capturing data from various instruments. Now you need to understand that in a time sharing system the task scheduler allocates time between various processes. In theory you could give every process the same slice of the pie but in practice you need to tweak things to ensure adequate disk read and write etc. How you tweak things is up to you but in a data capture situation, if the system is too busy with other things and doesn’t look at the instruments often enough it can miss spikes in the data. Obviously in a nuclear power plant this could be critical (bad pun).

To get Vaxes to be responsive for editing Dec artificially upped the priority of the terminal process, which meant that if a lot of people had active editing sessions open they could starve the other processes of time and you would start missing data - why? because the system was processing the letter A.

The answer of course is to treat data capture as high priority (or to down the terminal session priority a bit rather than just using the out of the box config.)

Incidentally we did late run a word processing system at York on Vaxes using WordPerfect and the enhanced terminal response feature made it usable but at the expense of a slow startup for login sessions. I’m going to guess that some of the latency problems seen with Roll.App is due to starting a process and all the virtualisation stuff that goes with it and getting things set up, the idea being that users will wait for a session to get going.

Except of course they don’t, increasingly people are used to fast responses, in part because a lot of desktop and tablet applications offer near instant gratification …

Thursday 9 January 2014

Well. my 3G router now works ...

As I’ve said before, I’ve invested in a 3G router as a possible alternative to our flaky ADSL service.

enter image description here

When I first set up the router it wouldn’t recognise the Huawei E173 3g modem or connect - it did know a modem was attached, but that’s all - and this is despite the fact that the unit is clearly labelled as supporting the E173.

I raised a support ticket with TP-Link who replied with a set of helpful links to their troubleshooting documentation.

Essentially, their solution was, inside the configuration tool, select the country to be Other rather than Australia and hand configure the settings.

VirginBroadband, my 3g supplier publish all these online as part of their BYO modem program, so it wasn’t a big ask.

The first problem I hit was that my router didn’t have an Other option in the country settings. To be fair, TP-link did suggest upgrading the firmware first and I hadn’t done that, principally because the unit was on a fairly recent version already.

So I thought laterally. I knew that the router explicitly supported the E173 for Turkcell in Turkey, so I reset the router, told it it was now on the shores of the Bosphorus, manually changed the APN settings to VirginBroadband, and to my utter surprise it just worked.

The modem connect light turned a solid cerise. No fanfare, just a solid data connection.

I must confess I was actually concentrating on my netbook setting up some other features, and didn’t notice the light go solid, but I only realised when Skype suddenly popped up to say that my old friend, Bill Parod at NorthWestern, was online.

Some more tests checking various newsites to confirm the dates and using whatsmyip.org to confirm I was connecting via Virgin, and we were away.

The router reconfigured itself at some stage to say it was in Australia and not Turkey but everything kept working.

My guess is that inside the router firmware is a table of countries, isp's and supported modems as part of the automated setup process. So if you select Australia you get the Australian cellular isp’s they support and the modems they use, and neither Virgin or the E173 was on the list. Forcing it to Turkcell let it correctly identify the modem, and hey presto, it could connect once I manually changed the APN settings.

So, onto stage 2 - I’ve 8Gb of postpaid data on the modem for testing purposes, so I guess the next thing is to transfer the ‘serious’ computers over to it to see how fast/reliable it is in practice …
Written with StackEdit.

Wednesday 8 January 2014

Nixnote and Everpad as Linux Evernote clients

One of the things that has always stopped me moving over to Linux full time has been the lack of a decent Evernote application, for these occasions when you are away from base but need to check up on a document.

I live inside of Evernote - it’s what I use to manage projects, file documents and the rest. And this means that there are occasions when the web client will simply not do.

I need the bells and whistles from a desktop client and the performance of a local copy of the database. While I’ve found accessing Evernote from the web client on my Chromebook usable, I do find I need to go and tidy things up from a desktop client.

For a long time I’ve used Nixnote on Ubuntu. And while I’ve had my differences with it over the years it’s now reasonably stable, but just a trifle slow. Certainly good enough for what I need for work.

However I recently came across a possible alternative - Everpad. Written in Python, installation looked to be straightforward, and certainly it did seem to work more or less out of the box. Except it didn’t.

scrrenshot of Everpad with crash

It bombed out during the synchronisation process. Which was kind of annoying. (Although to be fair, NixNote has done this to me in the past).

As a product I’d say it’s not quite there as a replacement for Nixnote. Also the interface is a little minimal - not necessarily a bad thing, but searching for notes (and I’ve around 5000 of them) is not implemented. It basically appears to provide no more than a local cache of the data - something for which there is a definite use case but it’s not my use case.

However, in spite of this I'd say it does have some uses - for example if Evernote was being used as a shared project notebook where the number of individual notes (and notebooks) within the database was not too great ...

Tuesday 7 January 2014

3G router fun

Over the weekend I meant to play with Roll.app on my Chromebook. I didn’t quite get there as I started playing with my new TP-Link 3G router.

When we were in Sri Lanka in mid 2013 I noticed that a few of the smaller places we stayed in used TP-Link 3G routers for their internet connections. These are essentially a wireless access point into which you plug a compatible 3G or 4G dongle, and then share the mobile broadband connection between clients, without the need for multiple dongles etc.

So, given that our ADSL connection runs over some 1980’s overhead copper phone lines via a neighbour’s apple tree, and is prone to attenuation and drop outs, not to mention that the adsl service is grossly overloaded with too high a contention rate it might be time for a technology change, especially as given our monthly internet usage we can get a 4G bundle that would give us 150% of our monthly data usage for the same price we’re paying for a not very good ADSL service (ie we use between 8 and 10GB a month and can get 13GB of 4G for what we are paying for 50GB ADSL we’re not using fully).

Now I had a Huawei E173 modem that we used on our South Australia trip back in 2012. I checked the modem specs and yes the E173 was compatible with the MR3220.

So, I did the lego thing. Plugged the WAN port of the 3G router into the current ADSL network box, plugged the 3G modem into the port, connected a laptop via ethernet for the initial configuration and away we went.

Rather than have it was a standalone network, I was going to setup the 3G router to use the ADSL connection as the primary and the 3G as a backup, and then once everything was working nicely move the 3G over to being a standalone network.

The idea being to see how good the 3G service was in practice by testing it fully, but leaving us with a mostly working alternative in the meantime.

The WAN configuration bit worked well, and was reasonably fast. The only problem is that it there isn’t a way to set a a low DHCP timeout - given our unstable ADSL and the time it takes to come back it’s possible to have the TP-Link think there’s no working ADSL connection - if it rerequested an ip address every hour or so that would probably solve that problem.

However the real problem turned out to be the 3G connection. When I plugged the 3G dongle in I got a flashing blue light as I’d expect. Unfortunately, while the TP-link box allegedly knows about E173 modems it doesn’t recognise my Virgin Broadband E173 modem - which is a bit of a problem.
In fact it’s a bit more complex than that - for Australia it doesn’t ‘know’ about Virgin, only Telstra, Optus and Vodafone. I did try Optus as Virgin effectively resells the Optus service, but that didn’t work, even following Virgin’s modem setup guides.

Looking round the TP-Link support site they had some E173 config files for TurkCell and UK Vodafone - I downloaded them to see if they were readily hackable, but they turned out to be binary:

enter image description here

Which was a problem. Googling didn’t help me - I did discover what the APN setting should be but without a modem configuration file I’m a bit stuck. I’ve a support ticket in with TP-Link at the moment …
Written with StackEdit.

Thursday 2 January 2014

Playing with Roll.app

Over the holidays I tweeted a link to a Document Foundation press release to the effect that you could now use Roll.app to use Libre Office from a Chromebook or an iPad.
Just to be different I tried it with both Firefox and the native Android browser on my seven inch tablet, and it certainly doesn’t work on Android. However it does work quite nicely with Chrome on a Mac

enter image description here

Essentially Roll.app is an application hosting service that lets you access an application via your browser. Libre Office is just one of a range of open source applications and tools. Storage is provided via Google Docs, Dropbox and the like.

In use Libre Office was fast and responsive and I have no reason to expect it would be any different of a chromebook.

They have a 14 day free trial period after which it’s USD0.99 per app per month, or if you want to put up with annoying video ads, you can use the free service after your 14 day trial is up.
The obvious use case is when you are working on a complicated document and a basic text editor would not do. It does of course also assume reasonable connectivity and bandwidth …
Written with StackEdit.