+--------------------------------------+
                    |           Welkin ToDo List           |
                    +--------------------------------------+



known bugs and limitations
==========================

 - graph layout algorithm is a brute force n-body, with algorithmical complexity
   of O(n^2), with n the number of nodes. There are algorithms that achieve at
   least O(n*log(n)) complexity, but they require a b-tree creation. We need to 
   investigate at how many nodes the nlog(n) algorithm starts to make sense and
   whether or not this is actually useful in real life due to data pollution.
   
 - all panes should work in concert. When you restrict the graph from one pane,
   all the others should update at the same time.
 
 - there is no 'usability degradation' control: welkin will not stop users from 
   loading an eccessive amount of information that would make the system too 
   slow to be responsive. we must think about ways to prevent this from 
   happening as usability should always be good.

 - long literals are not fragmented and flowed in nice text boxes.
 
 - hue color sliders are lame, we need to find a better method for color coding
   that would make it easier for users to select colors that have a cognitively
   similar level of contrast. (see the NASA usability lab color tool for an
   example of that).
 
 - fish-eye focusing looks sexy but might not be that effective in real life,
   maybe semantic zooming/panning would be more helpful? or combining the
   approaches? in case, when doing the fisheye zoom, we should make sure that
   things don't fall off the edge of the graph and get cropped.
    
 - add the ability to filter by type


in the queue (and in no particular order)
=========================================

 - allow models to be loaded from a URL (a web-browser-like address bar at the 
   top would be handy) or from a SPARQL web service.

 - add provenance info to the graph model (basically, being able to distinguish
   groups of nodes by where [what file, what web service] they came from)
   
 - allow the current model + current layout to be saved for later reuse.
 
 - play with clustering algorithms as a way to simplify the graph when it 
   contains too many nodes (Newmann-Girman is very promising, but it's 
   algorithmical complexity may be somewhat limiting and clearly not realtime, 
   but maybe we can reuse some of the force-repulsion simulation to perform the 
   clustering incrementally)
   
   
   
              ===================== WARNING =========================
              
   We are perfectly aware that graph-based layouts are not meant to be for 
   general users (ie. our grandmas) and exhibit a very low efficiency in 
   finding individual items. Welkin is not a browser you use to get to the 
   single item, but more one that you want to use when you have no idea about 
   the data you are looking at. Sort of a "data telescope" if you wish, where 
   the clusters are more important than the single nodes. So, no, we don't
   consider this a bug or a limitation, even if we do keep data pollution and
   clutter minimization a priority in our UI research.
   
            ==========================================================