Sunday, 26 August 2007

A Matter of Reliability

What do you learn in a community-building exercise when the community building almost fails to happen?

One obvious answer is "Ahh, forget it. We are none of us perfect. Try again and it'll be better". Another is "Well, you guys have failed sometimes; cut some slack to the others". Both are fair answers. But the question itself, and the quality of the answers, underpin a more crucial aspect of the governance of 'bottom up' SDI's, namely how do we fare when, later on, there are services over which we've built our own value-adding services and, tomorrow, the service custodian goes out of business, so to speak: a change of policy, a shift in budgets, loss of key personnel. Any number of reasons might pertain, but suddenly our customers are no longer happily receiving their service.

The point is that when you expose open-standards services to the web then I can come along and build on your service, adding a new value or servicing a new audiences that you hadn't planned for. Sure, you can argue that I'd be at least slightly daft to build a critical need on your service without some sort of agreement, let alone recognition. Okay, but, when I try to make a phone call to the other side of the world, my success or failure depends on a whole chain of agreement between telco operators, any one of which can fail just when I need the service. To what degree can I blame my local telco? Not much, if the failure is three networks away in Outer Mongolia. Problem is that it doesn't matter to me where the failure is - my call hasn't gone through. I'm an unhappy frustrated customer.

So, Tuesday should have been a neat day - meeting with the secretariat of the Kenyan national SDI, with two purposes were in mind, Firstly, helping them set up for evaluation some open-standards tools for establishing a national metadata archive, and the ability to serve geo-spatial data directly to remote clients (Geonetwork and geoserver, respectively). Secondly - and more importantly for me - engaging KNSDI's help to organise a meeting in September where we hope to get together all the national SDI players in East Africa along with their UN counterparts. The purpose? Try to start mapping the institutional interfaces that will be needed between the national and regional SDIs and the emerging UN spatial data infrastructure counterpart.

The UN is meant to serve member states, and the UN relies upon member states to provide data and services needed to inform and address trans-national and global issues. If a UNSDI's purpose is (amongst other things) to promote interoperability, shall this be on the basis of 'best effort' by the member states? Is the UN obliged to help members meet minimum levels of reliability and accountability? If so, what levels, and who is to measure and ensure them? If not, what is tolerably good enough, and what happens when gaps in data availability or reliability lead to flawed assessments or decisions? These and many more questions require attention, probably over and over as methods and approaches are tried.

So what happened? The Nairobi traffic Gods frowned, and three out of the four KNSDI participants got stuck in a jam. Could happen anywhere, couldn't it? Could as easily have been a delayed flight, a flood, or sick child to deal with. You bet, all very normal and we can schedule around it and (I hope) we can get our regional meeting organized notwithstanding.

But it did seem to me a specially pertinent reminder that, as we in SDI-EA try to promote SDI and build interoperability amongst distributed services, that the fragility of the communications and transportation infrastructures remain a constraint. What does it mean if we can implement a world-class on-line repository for satellite images in Nairobi if operational agencies and NGOs cannot access it precisely when needed? Who is liable if, in three or five years' time, humanitarian services' crucial decisions are delayed by lack of reliable data services.

Tough questions that won't go away if we ignore them. Neither, it seems to me, are they likely to be more easily answered unless we start now to articulate and specify the requirements to which telcos and other service providers can respond on a fair and contractual basis.

Anyway, we were back to KNSDI on Wednesday, had a happy and fruitful meeting to organize the consultation, and John at least got through the Geoserver installation training. We'll still need more time to get Geonetwork in and running for them, but at least we hit this important milestone and - fingers crossed - the KNSDI folk will like it enough to move it across to their production server. I'll be especially pleased if we can go into the September workshop with Survey of Kenya delivering on-line one of their signature framework data layers, like administrative boundaries. That really would be a feather in the collective caps. But only if it can be made reliable.

Sunday, 12 August 2007

Teflon rules!

Friday two weeks back was the day I've been waiting for - the day the Teflon Principle seemed to actually kick in.

John and I were invited to join Byron from RCMRD ( in a trip to meet the staff from the Geomatics Unit at the Jomo Kenyatta University for Agriculture and Technology ( Professor Gacahri out there is one of the movers in the KNSDI, and had responded to Byron's suggestion that some of the staff out there be given a walk-through on what, in practice, participating in operational SDI could mean to them.

The first surprise was thelab facilities. Remember, I've been in East Africa since the times when a simple PC cost a lecturer's annual salary, a single diskette cost 10 dollars, and too many undergraduate's use of GIS was limited to what they could read about in text books. Here were 30-odd PCs, a functioning LAN and good internet connectivity. My, how things change. And not a bad start at all.. no server, though.

During discussions two interesting things emereged: the first being that, when they'd had the Geonetwork toolkit described to them, the JKUAT staff immediately saw its usefulness to teh school as a potential publisher and provider of geospetial information (as well as the - to me - more obvious attraction of being able to find stuff). The second, and the one that always makes John happiest, is when they twigged that establishing and managing a geospatial would not only streamline their data provision to the students but would open up the possibility for lateral integration of data across studies, rather than just vertically within them. John's alway happiest when this data cohesion aspact emerges spontaneously.

So, the followup is that somewhere towards end of August John and Byron and I will be back out to JKUAT to do the full-blown hands-on training with the Geomatic facluty, and then two weeks after that to do another with the final-year students, though for that one I'll push that it should be the faculty that do it, with John and other's back-stopping them. Like I say, we have to maintain the Teflon Principle, and the idea of a centre of excellence emerging at a school that's already leading the effort for geospatial awareness in the region, achieveing it without any massive financial outlay, and doing it collaboratively with the RCMRD and their training services seems like a marriage made in heaven.

Frustrated Ambitions

One purpose of attempting SDI-EA has been (optimistically?) to test how feasible SDI technologies are outside OECD-type countries, using the UN's own capacities to augment local infrastructure where needs be. Over the last two week has been an exercise in frustrated realization of what the limits can be like.

It all started with a flurry of requests for data and land cover change analysis - one from the GEF evaluation office, the second from a UNEP study of refugee camps, the third from researchers in land use conflict avoidance in Kenya and Tanzania. All good stuff and, thinks I, a good chance to test some of my theories against cold, hard reality.

All these requesters, by thge way, were suffering from the delusion that UNEP/GRID still acted as some sort of massive data archive that would have the necessary data on hand. Alas, no. That's a business model that went the way of the dodo many years ago. Frustration #1 was discovering that our own backyard is in desperate need of a cleanup - the Landsat data and stuff that I know NASA delivered to us years ago is nowhere to be found. Or, more accurately, no-one knows where to find it. Oh dear.

Anyway, this is the age of the internet and postals and all we need do is know how to find the data and use our satellite capacity to pipe it into Kenya for our clients, right? Theory says that we can be clever and use on-line services to slice out justthe bits of the images for relevant study areas - a few megabytes rather than 10's or hundreds of them, smart use of limited bandwidth, more readily accesible to users in developing countries and so on.

Second frustration: NASA's geobrain ( usually provides a neat web coverage service whereby you designate your are of interest and it goes off, interrogates the LAITS catalogues and comes back with thumbnails of the available scenes; you make your selection and it then goes to the WCS, excises the footprint you've selected and send a nicely bundled tar package your way in a matter of minutes. What's wrong with this picture? Just the fact that the services is off the air this week. Sigh

So now I'm using sites that only deliver full scenes, such as the Global Land Cover Facility's Earth Science Data Interface ( Obviously a much greater demand on our satellite link but worth a try. Except for the fact that the link has been slow and flakey and up and down all week - what might otherwise be an easy 20 minute 16 Mb transfer sometimes taking half a day. Sigh. Not the sort of reliable service we'd like to offer our partners.

Glad I didn't offer to mortgage the house as guarantee of being able to deliver on the requests made to us. Maybe things will be better next week. It's still not a great advertisement for SDI, is it?

------------------------ 24 Hours later ---------------------------
Well, things have picked up a bit, and last night I actually managed to pull down 4 MMS scenes and a full TM set, a total of about 350 Mb - not a huge volume of data in these days of streaming media but significant in this part of the world. Now, if only GeoBrain start behaving itself....

--------------- ... and 24 Hours after that? ----------------------
Well, some hacks and work-arounds later, I've managed to pull down about 10 Thematic Mapper and half a dozen MSS scenes, a total of about 2 gigabytes of compressed data moved as scheduled downloads overnight while UNEP's bandwidth is mostly unused. Frustration the Thjird:As it turns out, the usefulness of all this data was pretty limited, usually because the change signals being sought were not significant at the resolution of the Landsat data. This is where having had Geobrain working earlier would have really helped: at least the limitations of the images would have been apparent earlier and more quickly, and the requesters could have adapted their expectations.

Now, this gets me thinking: why am I downloading data from Maryland anyway? What if the Regional Centre for Mapping Resource Development here in Nairobi could bring its Landsat archive on-line, a sort of Geobrain East Africa? Must have a word with my friend Byron about that possibility...