The Ndiyo Blog

Geeky Linux post - you have been warned!

Those of you who create Live CDs for Linux probably know about UnionFS, but others may not. Basically, it’s a filesystem which merges the contents of one or more underlying filesystems. Some may be read-only, others read-write. The filesystems can be from different disks, different media, different servers.

Its most common use is to make a CDROM-based filesystem ‘writable’, by putting a read-write ‘overlay’ on top of it, either in the form of hard-disk directory or a RAM-based filesystem. Any changes to the filesystem get written back to the overlay. This could be a good way to build your own variant of a standard distribution CD before burning it.

There are some other nice things you can do with it, though.
(more…)

We had a few comments from people suggesting that it was too difficult, on the Ndiyo web site, to get a feel for what the project is about. At least, to do so quickly.

So we’ve done an Executive Summary which gives you the big picture in 12 easy points!

One development which makes the Ndiyo model ever more viable is the growth in web-based applications.

Whenever I want a map these days I turn to a web-based service, where once I would have used Autoroute or Street Atlas on my PC. I posted a message to a newsgroup this morning, but I used a web interface and not an NNTP newsreader program. I don’t, in general, use webmail on a daily basis, but my use of it is becoming more and more frequent. Writely is a great way to collaborate on documents, Google Calendar is looking very good, and so on.

Why is this important for Ndiyo? Because it is becoming less and less relevant which operating system you’re running, and more and more relevant which browser you use. As I wrote elsewhere recently: “There will be a viable Open Source desktop environment after all. And it will be called Firefox”.

This means that systems like Ndiyo’s can take advantage of the benefits of Linux - stability, affordability, security, multi-user support - and not be as handicapped as they once were by the lack of friendly desktop apps, or the possible unfamilarity of the apps on a particular platform.

Ismael Chang Ghalimi coined the wonderful phrase ‘Office 2.0′ and has been writing about his experiences using all web-based apps on his IT Redux blog. He’s building databases of Office 2.0 products, documenting his own setup and so forth.

Recommended reading.

Many thanks to Mark M Smith for the link.

One of the lessons we’ve learned from our trial deployments is that the level of Free or Open Source Software (FOSS) expertise that is available locally can be a critical factor in determining the success of a system. Such expertise is widely available and widely distributed in the developed world, but appears to be unevenly distributed in developing countries. At the moment, though, data about this seem to be mainly anecdotal. So it’s interesting to see that Bridges.org, an international organization with a mission “to promote the effective use of information and communications technology (ICT) in the developing world for meaningful purposes”, has just completed a substantial survey of the position in several African countries. It makes interesting reading. Key findings include:

Both free/open source software and proprietary software can be used to offer technology solutions appropriate for African public-access computer labs.

The thin-client model provides a reliable, cost-effective and popular solution for public-access computer labs in Africa.

Software license costs for proprietary software are significant in principle, but in practice they are not borne by many of the public-access computer labs in Africa.**

At ground level in Africa, the potential for cost savings gained from the use of FOSS depends on many factors.

General ICT skills levels — especially for installation and maintenance of software — are low overall, and experience with proprietary software is more pervasive.

Training courses for PS are more widely available than for FOSS.

The fact that FOSS makes source code available and encourages modifications is not exploited by the vast majority of public-access lab staff or users in Africa because they lack the necessary skills.

However, it does offer an opportunity for local service providers to create customised applications.

The availability and quality of local technical support in Africa (for both free/open source and proprietary software) is reasonably high overall, although FOSS support tends to rely on free services.

While there are a number of projects underway to translate software into African languages, these localised versions are not yet widely used in public-access labs and there is some disagreement about the value of local language software.

Most FOSS labs in Africa are set up and supported by a small group of enthusiastic implementing organisations, so the success of FOSS labs relies heavily on their efforts.

General trends with regard to economic situation, age of the lab, and staff experience can be identified among labs that use the same type of software.

** Note: Absence of licensing costs for proprietary software is explained by the study in these terms: either the software is donated; or the software is pirated. While the second explanation may be accurate at the moment, it doesn’t seem to us to be a sustainable reason for not counting the cost of proprietary licensing regimes. You can’t build a sustainable ICT base on piracy.

At Ndiyo, as everywhere else, we are governed by Hofstadter’s Law: It always takes longer than you think, even if you take into account Hofstadter’s law.

The most common question we get asked is “How/when can I get hold of the Ndiyo hardware?”. Oddly enough, it’s the hardest one to answer. Let me explain…

The ‘ultra-thin-terminal’ device we use is based on the Nivo technology developed by Newnham Research - a company we created some time ago, but which is independent of Ndiyo. Newnham uses the technology in a variety of products, and the Ndiyo ‘ultra-thin-terminal’ devices for Linux have not yet been ‘productized’ for mass-production. Devices ordered by Ndiyo for trial deployments are specially built and take about 3 months from order to delivery.

Because of our historical relationship with Newnham, we’ve been able to get access to this technology early to perform trials so we can test it out in the real world and show that it works. But before we can start shipping things to ‘customers’, even trial units, there’s quite a lot to be done:

• We need to build an e-commerce and distribution channel, or partner with somebody who has one
• We need to create the support infrastructure and online community to deal with any user issues that may arise
• We need to do some software development to make the systems easier to support, maintain and use
• We need to estimate the market size and decide how much stock to order
• We need to hire the people to do the above - Ndiyo is a very small organisation at the moment
• We need to find the funding to do all of this

It’s no good having a great idea like this and not being able to deliver it to the people who want it, and it’s especially frustrating when the technology is so close to being ready. Also, it’s no good shipping units and not having the resources to support users. So we’ll be taking a break from trial deployments for a while, and over the next few months we’ll be working as hard as we can to get these other components in place.

Sigh… if only hardware was more like software!

In the meantime, there are lots of interesting developments and ideas that we also want to keep our eye on. I’ll be posting more about those in due course…

I’d like to share a picture with you, taken just now in the Ndiyo office.

It shows my MacBook Pro, and next to it is one of our Nivo-based terminals connected to an Ubuntu Linux machine. Why am I showing you this? Because the Ubuntu machine is not a real machine. It’s a virtual machine running on the Mac. (There’s a console window for it in the top-left of the Mac screen). The performance of the terminal is remarkably good.  I cropped the image above using Photoshop on the Mac, uploaded it to the blog from there, and then, just for fun turned to the terminal on the right and I’m writing the rest of the post using the Linux version of Firefox.

Virtual machines are a hot topic at the moment, but we’ve been watching them particularly carefully at Ndiyo, because when you combine them with low-cost thin clients, you can do some quite interesting things. Not only could my Mac be running several Ndiyo terminals from this Linux virtual machine, it could also be running Windows virtual machines and providing Windows terminals.  VMs are not a very efficient way of sharing computing resources, but they’re a pretty flexible one!

Our main interest at Ndiyo has been that a single Linux server in a small business could provide, say, five Linux terminals and one Windows one (for the person who needs to run the accounts package). Quite cute…

I found a problem with the narration soundtrack on the South Africa movies. If you tried downloading them and found the narration difficult to hear, please try again - it should be much better. The video encoding is improved, too.

For the technically curious, my software had, for some reason, recorded the narration as a stereo track where each half had an opposing polarity. On systems with good stereo separation it sounded a little strange but was basically OK. On others there was a tendency for it to cancel itself out. I flipped the polarity of one track and it fixed it.

Yama Ploskonka wrote in:

Yes, the blog is a great idea.

Then I went back to the main pages, and saw again the Nivo. You have an extra connection for power-in - which often means a dedicated, energy-non-efficient power transformer for each, one more wire, etc etc.

I’ve been playing with using the usually unused ethernet wires for power. Turns out there’s even a word for that, power-over-ethernet.
You use normal ethernet twisted pair cables - no need for anything fancy. I would be happy to help design the (small) box that ups the necessary DC to the wires. Obviously the Nivo would require either a redesign. or a down power module. All in all, cheaper than the power transformer.

Thanks, Yama - that’s a great offer - and you’re not the first person to ask about power-over-ethernet. It’s true, we could use it, and if we did it would probably make sense to use the standards that are now in place. It would also help to reduce the tangle of wiring that typically accompanies an Ndiyo installation at present!

There are a few reasons, though, why we haven’t gone down this route yet:

• Standard PSUs like the ones we use are really very cheap, typically less than 5 dollars even in modest quantities.
• Network switches which incorporate power-over-ethernet tend to be substantially more expensive than those without - much greater than the cost of the PSUs. And then you need to add the components to extract the power at the other end.
• We could build our own non-standard system, but we’d have to warn people about not plugging it into conventional switches or hubs
• In either case we’d need to consider in advance how many Nivos we wanted to power, and sometimes we’d end up with unused components or over-specified power sources because we weren’t using the full capacity
• We have to have a source of power at the terminals to drive the monitors anyway
• There would be a fair amount of cost in redesigning the Nivos

It’s something we should keep looking at, though. And there are some other interesting options. Quite a lot of flat-panel displays incorporate USB hubs. If that trend continues, we could run the Nivo using the power available from USB.

The other option is to do powerline networking - ie. ethernet over the power lines, rather than the other way around. You can then get rid of some of the network cabling. The bandwidth of these systems is starting to get fast enough, but at the moment it’s still an expensive option and has regulatory problems in many countries.

Certainly something we’ll keep watching, though!

Virtually since Ndiyo started, we’ve had a fruitful relationship with Cambridge’s Judge Business School, and particularly with Bill Nuttall, who runs the School’s Masters in Technology Policy. Bill has been very helpful in pointing some of his graduate students in our direction. The deal is simple: they need real-life problems for their dissertation research; Ndiyo has real-world problems we’d like to see solved — or at least explored in depth. So students come and work with us for a few months, and in the process we all learn from one another. Ebenezer Ademisoye — who starts tomorrow — is the third graduate student from the Technology Policy programme to work with us, and we’re delighted to have him follow in the footsteps of Amit Srivastava and Imran Ahmad.

For most of the deployments that we’ve done so far, we’ve configured the terminals to log in automatically as ‘user1′, ‘user2′ etc after a short period of time. The systems therefore provide walk-up web browsing but don’t allow return visitors to come back to their bookmarks, their emails etc.

We’ve been trying to find Linux software which provides user-management for these kind of situations. Ideally it would also monitor the duration of a particular user’s login for billing purposes, simplify the creation and deletion of users, and so forth.

There’s the Zeiberbude software, but it doesn’t look as if it’s been updated for some time. DireqCafe looks more hopeful. OpenKiosk looks even better, but is dependent on running a KDE desktop; we’ve been using the default GNOME-based one that come swith the standard Ubuntu. KDE may be a sensible choice anyway - I’ve heard that it’s easier to lock down and it may be more familiar to Windows users. But if there were something that could be made to work with Gnome it would be very convenient.

In the Community Phone Shops where we did our most recent installations, the owner provides a set of payphones connected to the mobile phone network. A set of booths in the shop have phones in them with simple control pads. (Click the images for larger versions)

Behind the counter, the owner has control of the main part of each phone and can keep an eye on the amount of time used, payment due, etc:

There’s now an Ndiyo server behind the counter as well, and the Ndiyo architecture could allow a very nice equivalent to this familiar phone-based system to be built for internet access… it could probably even be built to have the same menu layout as the yellow control boxes to make it really easy for the users.

Anyway, if you have any experience of these or any similar packages, please let us know about the pros and cons!