Impressions of the 15th Generalisation Workshop
So I participated to the 15th Workshop of the ICA Commision on Generalisation and Multiple Representation and it was really inspiring for me. As there were so many critics, hints and new ideas and I need to collect the basic impressions! Therefore I will write here a conclusion from my point of view.
Ontological intentions and relation modelling
The workshop began by some (for me) very interesting efforts on ontological approaches. Especially Nick Goulds presentation on “An Ontological approach to On‐demand Mapping” introduced (at least for me) the ideas of Web Ontology Services and Web Reasoning Services. As my own efforts are targeted on the integration of Web Generalisation Services to the automatic computation of interactive tile-based maps, I think that the presented concepts could represent the missing link in the process chain for applying web services to automatic on demand maps. Also the presentation on “Towards an Ontology of Generalisation Constraints and Spatial Relations“, held by Sandrine Balley, concluded many ideas and facts that will surely (or even have to) flow into my further investigations.
When I was working on my tool for the automatic integration of WebGen-WPS to tile-rendering, I did very often realise that it is really hard to implement an automatic detection or evaluation of a feature types demand on generalisation. That is essentially for the automatic appliance of generalisation operators that need to take the spatial relations of a feature into account (e.g. displacement, aggregation). Also, there is a kind of a gap between the pure indication on the need for the appliance of a generalisation operator and the related practical appliance of functionality (respectively algorithm) in the fully automated process flow of rendering (computing) map tiles. Therefore, the modelling of appropriate ontologies is a concept that could really help to bridge the gap. I would like to try using these approaches and I am very excited on further investigations on these ontological modelling efforts.
Web Generalisation Services and Moving Code
Stefan Wiemann introduced, in his presentation on “A framework for building multi‐representation layers from OpenStreetMap data“, the term Moving Code that noticeable left a lasting impression to the most participants of the workshop. The related concept provides the possibility for exchanging algorithm & functionality and could have (in my opinion) a clear impact on the development of generalisation functionality in the future. For me it did already influence the way I envision the further developments of WebGen-WPS. I always had (and still have) the vision of using this framework to give the open mapping community of the Web2.0 (applying ‘Neocartography’) the opportunity to:
- use generalisation functionality (algorithms)
- develop own generalisation functionality (algorithms)
- share functionality with each other
- let community test developed functionality (algorithms)
In my opinion, the scientific generalisation community as well as the ‘neocartography community’ would benefit from such a platform.
WebGen-WPS is currently a very static opportunity to apply generalisation functionality via WPS. It was a first approach for providing generalisation functionality in an interoperable form but has not arrived an atomic applicable status. I think the biggest problem is, that a user can just apply the algorithms which exist. The concept of Moving Code led me to the opinion that Web Generalisation Services have to be way more interactive and need to involve the user as much as possible. But which advantages would a user have to develop an algorithm and share it with other map makers?
Objections and Problems
I referred in my presentation to tile-based mapping and how I think it would be possible to use the WebGen-WPS in its current status. I also spoke out the idea on the corporate development of generalisation functionality via Web Generalisation Services, described in the previous section. Legitimately some objections and problems where argued which I want to conclude now a little bit.
Exceptions came especially from participants which represented NMA’s and they gave me good input for my further considerations on which aspects have to be observed for implementing a broad accepted and applied Web Generalisation Service for a corporate development of algorithm.
Basically there are five different groups that apply cartographic generalisation and could/would generally participate/benefit from corporation:
- national mapping agencies
- commercial map providers
- scientific researchers
- community driven mapping initiatives
- private mappers
These different groups can use…
- commercial software
- open source software
…to apply generalisation and develop appropriate functionality (algorithm).
During different discussions I noticed that there are problems related to licenses when these different groups would really work together. For example, when a NMA uses ArcGIS for developing an algorithm and it uses ESRI-functions for realising a generalisation operator. It is not possible to share this code with a community as there are proprietary objects and methods in the code. Or if a researcher is using some proprietary data and wants to process geometries external at the WebGen-WPS, he/she cannot be sure what happens to the data. Will it be latched and used for anything on what the researcher has no more influence?
Also NMAs see a general problem in using open source software as they cannot really trust in it. Is it really stable? Will the development stop in one year? Or even…where do I get support when I have a problem?
These exceptions indicate that a collaborative work between the 5 groups is really hard to realise…but that doesn’t mean that it is impossible. At least the last 3 groups are easily able to develop code in collaboration.
To prevent the doubts on the licensing I think that it is necessary to ensure that a Web Generalisation Service, that builds on the collaborative development of algorithms, offers the applicants different license-strategies for his code. Whereby different kinds of collaboration/participation are possible. The license for the source code of an algorithm could be defined in four levels:
- downloadable & completely changeable
- downloadable & visible but unchangeable
- not downloadable & visible but unchangeable
- not downloadable & invisible
(At this point I just save my thoughts and do not care about existing licensing strategies. I will do that later!!!)
The first and second points imply the exchange of code via moving code and for the third and fourth point the related functionality would be just offered via WebGen-WPS.
In one break out session it also became clear that adequate parametrization strategies are very necessary. Especially for simple embedding of functionality into an existing process chain, it has to be exactly predefined which parameters are initially needed and which will come out at the end. For the appliance of WPS this is already offered by the DescribeProcess-request, but it is even more necessary when code will be integrated locally after loading it.
One note was also, that a kind of “effect-matrix” of the algorithm could also be helpful, which gives a kind of preview of the effects that the appliance of a functionality will imply.
After the fourth session, on “workflow and production systems for topographic data”, where especially NMAs presented their progresses, William Mackaness floated that all the presented concepts and workflows are working on so similar problems and each one seems to invent the wheel again. Maybe inspired by the ideas and concepts that Stefan Wiemann and I already presented…maybe just because it really all seemed to be relatively similar, he asked if there should not be more collaboration and exchange between the single NMAs. By that I realised, that there are three different levels of possible collaboration:
- sharing implemented code (algorithms)
- sharing concrete concepts
- sharing pure information
These are not just related to the collaboration of NMAs…they are also applicable on the possible collaborations between all 5 groups that I have previously defined.
Sharing concepts and information is already applied by participating on workshops and conferences. But that could be essentially boosted by much more using the advantages of Web2.0, meaning forums, mailing lists, wikis, etc.
Sharing code would be a completely new field, that could essentially prevent from having finally numerous implementations of one generalisation operator, instead of using this energy for collaboratively inventing a completely automatic mapping solution which still does not exist.