[MOBY-dev] Re: [unclassified] graphs again...

[Mark Wilkinson]

getting closer every time :-)

There are still some peculiarities...  it seems to revolve around a 
specific interpretation of service namespace.  IMO - a service that 
registers itself as consuming * object in the N namespace can only be 
fed by a service that produces an object in the N namespace (not by a 
service that produces objects  in * namespace).

So, for example, there should not be a connection between 
getGoTermAssociations and SearchPubmed, because the former outputs 
*unknown* namespace, but the latter *only* consumes the global_keyword 
namespace, so it is unlikely (though possible, I suppose) that these 
services will feed into each other.  I think that we should assume that 
they cannot.

At least, that's the way I view the world.  I think that connections 
that are not true for all cases should be left out of the graph.

M

Martin Senger wrote:

>So here we go. The next version of the services graph is available as:
>
>http://industry.ebi.ac.uk/~senger/jMoby/images/services-LR.png and
>http://industry.ebi.ac.uk/~senger/jMoby/images/services-TB.png
>
>(Both images represent the same graph, just once the orientation is
>Top-Bottom, and the second image is oriented Left-Right.)
>
>The images are still quite big. If you have suggestions how to make them
>more readable (for example by using some colors) please let me know.
>
>But, more importantly, I would like to know if thet are correct, or if
>they still have too many or too little edges.
>
>Just fyi, the edges that connect collections with simple types, and
>viceversa, will be displayed as dotten (or dashed) in the next version
>(unless you suggest to remove them completely, or let them there only as
>an option, for example).
>
>Btw, the explanation what the arrow symbols mean is here:
>http://industry.ebi.ac.uk/~senger/jMoby/images/legend.png.
>
>Thanks for your patience.
>Martin
>
>PS, Yes, Rebecca's example is now connected correctly I hope.,
>M.
>
>  
>