[MOBY-dev] Re: [unclassified] more on graphs

Mark Wilkinson markw at illuminae.com
Wed Oct 15 11:52:48 EDT 2003


Martin Senger wrote:

>   http://industry.ebi.ac.uk/~senger/jMoby/images/data-types-newer.png
>
nice!  

>   The services graph looks worse. That's why I am asking if it is
>correct, or not. 
>
it is not.

>The main problem is with services getGoTermAssociations
>and PBI_Cluster_Blast because they do not specify any namespaces for their
>input and output. Therefore they are connected with many services that
>have just 'Object'.
>
If you look at the service signature for PBI_Cluster_Blast, it says that 
it consumes GenericSequence objects, thus it cannot consume 
VirtualSequence or Object objects, as your graph is suggesting...


>   * if a service claims that it has an input of type "X" but does not
>define any namespace for it, then any service with output of type "X",
>even if specified with some namespaces, can connect to it?
>
this is correct.  Absense of namespace indicates that namespace is 
irrelevant to the service.

>   * or the true is rather that services with a missing namespace in their
>input/output data can matched again only with services with missing
>namespaces (with the corresponding data types, of course)?
>
this is not true.

>   BTW, it's a pity that the API does not support yet the retrieving of
>the full ServiceType object. If you add it there (as mentioned for the
>future in the API documentation) I can easily add yet another graph
>displaying the tree of service types.
>  
>
I will add it as soon as I get a free finger to type with :-)

cheers!  Those graphs are VERY cool!  (even though the "Hello PlaNet" 
service still exists... hint hint hint!)

M




More information about the MOBY-dev mailing list