What's your view on the direction of GIS?

If you are in the data creation business, I don't think that there is any substitute for Desktop GIS. The limiting factor in these cases, is the large sizes of data, which would take an inordinate time over any kind of network.

The strength of Cloud based GIS is when you have a centralised server, serving out data, and viewing and limited editing functionality is required on the client side.

I think it is going to move from desktop to cloud, and here's why. I currently run virtual training courses where the users log on to Amazon virtual machines to do the exercises. Similarly, if I want to do some testing I just fire up a machine and get to work. It's simple, it's configurable, and it's efficient; and I think it's going to soon reach the tipping point where it's just natural to work this way.

My point is not to think of cloud computing as just a way to store and serve data: you can create virtual machines in the cloud. When you say a GIS user could have a lot of different uses and tasks, then the ability to use pre-configured machines like this means you don't need to have a desktop computer running everything. And the cloud is scalable, so for processing you just fire up as many cores/engines as you need to get the job done in the optimum time.

As for data, vector is certainly storable in the cloud right now. I think raster is feasible, though complicated by having to tile/pyramid data. Point Clouds (Lidar) is going to need an increase in bandwidth to be practical, but I can see that coming.

Cloud vs desktop represents a false dichotomy.

On the desktop it is common to access resources over a network or the internet. WMS, WFS, SQL and even file servers are all essential to a typical desktop GIS setup. Desktop GIS would be much poorer without the "cloud".

Data stored and processed in the cloud still needs to be rendered on a client machine. Client libraries such as OpenLayers are very powerful in their own right in terms of basic GIS functionality. Cloud GIS needs high quality client side code.

This dichotomy represents a hurdle to future development of GIS. To overcome it we need to change how we think about storing and processing data:

Storing Data

The current practice is to treat data as either local or remote. Either a shapefile on your system or a dataset on a server that is downloaded as you need it. Instead the data should be in the cloud by default, and cached on the client as needed without any user action. Network Links in Google Earth already allow for this in a very basic way.

The development of GIS specific version control systems akin to Git and Github also hold promise.

Processing Data

The current approach is to perform analysis in isolation. Data is brought into a system from the outside and processed. In contrast the Google Earth Engine processes data where it is stored; reducing storage requirements and bandwidth.

Despite the obvious advantages of cloud processing there is still a need for desktop tools. It is simply much cheaper to store moderate amounts of data on a desktop machine compared with current cloud offerings.

A possible future direction is developing protocols that allow data to be processed at the optimal location without user decision or intervention.

Conclusion

• The cloud is already an intrinsic part of desktop GIS, and client side code will always be essential to cloud systems.
• No single tool has all functionality, and the advent of the cloud does not change that.
• The future direction of GIS could be towards the development of protocols that allow the differences between cloud and desktop to be ignored.