[MOBY] [MOBY-dev] Moby Graphs version 1.0 (or 0.9?)

[Mark Wilkinson]

Hey Martin, 

I'm having a lot of fun playing with these graphs :-)

I've noticed a couple of things that strike me as odd:

1)  build a graph from GenbankAccSequenceRetrieve to PBI_Parse_Blast. 
If you follow the path from GenbankAccSequenceRetrieve to
GenbankAccVirtualSequenceRetrieve to GenbankCompleteSequenceRetrieve you
will notice that the first two don't feedback on each other, but the
GenbankCompleteSequenceRetrieve does.  I think all three of them should
feedback on each other (though in the context of finding a path the
correct behaviour is probably to suppress this)... so I guess the
question is why is it that some services are shown to feed-back, and
others, don't even though they could?

2)  On your interface page, you have the option to show paths between
selected services, where you have the service names in the drop-down
menu.  Be careful with this, because service name alone isn't sufficient
to specify a service.  The unique key is authURI + serviceName.

What's already striking to me is how many paths I know exist, but are
not shown because (a) we don't specify in the registry what
Cross-references a service provides, or (b) the output namespace isn't
unknown until run-time (e.g. the getGoTermAssociations should, in fact,
be able to lead to one or more sequence retrievals because what comes
out are usually EMBL ID's or some such thing).  This is not a criticism
of the graphing, but a criticism of the somewhat impoverished
representation of service outputs in the registry... but I don't see a
way around that because it is genuinely the case that sometimes you
don't know what you are going to get until you run the service.

Anyway, it's great fun!  Thank you again for spending the time to build
this!

I wonder if, rather than a service -> service graph, a more useful graph
might be from input data-type to output data-type?  I only say "more
useful" because service names are "officially" meant to be opaque, so
the drop-down list of services might not always make as much sense as it
currently does when all of our services are "sensibly" named.  Is that a
more difficult query to execute?

In any case, this rocks!  Thanks so much for your efforts - I really
should repay you by writing that DUMP method that you asked me for last
week... you're still too hairy to kiss ;-)

Cheers!

M

On Sat, 2003-11-08 at 19:59, Martin Senger wrote:
> Mark, Ken and others,
>    After some struggling (with firewalls etc.) I hope to have a reasonably
> ready version of Moby Graphs. The main page is here:
> 
> http://www.ebi.ac.uk/collab/mygrid/service2/jmoby/index.html
> 
> This page does a bit of explaining - and it should link you to the
> servlet-based service where you can play with creating various graphs
> (http://www.ebi.ac.uk/collab/mygrid/service2/jmoby/graphs).
> 
> I will be glad to hear your comments. Mark, let me know when is the time
> to announce it to moby-l (or do it yourself if you wish because I will be
> travelling the whole week and not sure how I will be connected).
> 
> Regards,
> Martin
-- 
Mark Wilkinson <markw at illuminae.com>
Illuminae