[MOBY-dev] BioMOBY Asynchronous Service Call Proposal v2.2 - <wsa:Action> tag missing for GetResourceProperty requests

[Tom Oinn]

Mark Wilkinson wrote:
> v.v. the political statement:  I strongly agree with Martin, in part, and  
> somewhat agree with him in another part :-)
> 
> My personal opinion on the Pseudo-Web Services approach that MOBY uses is  
> that it has been more of a barrier than a benefit.  I don't think this is  
> a consequence of MOBY not using all of SOAP/WSDL, I feel it is because the  
> WS architecture itself is not quite as useful as it was marketed to be in  
> past years.  In this regard, moving toward a REST-style architecture for  
> "MOBY 2" is something I am strongly in favour of because it seems like the  
> right thing to do, especially in the emergent Semantic Web world; the fact  
> that it also moves us closer to the S-MOBY architecture is just icing :-)

One thing to consider here is that you're potentially missing out on the 
additional capabilities you can inherit from a web service platform. I 
tend to agree with both you and Martin on this (Martin has sat next to 
me in an office for some years and so should know my generally low 
opinion of the current state of the web service world!).

Where WS can really come into their own is their ability to add aspects 
such as security and robust message transport without any additional 
effort from the MOBY community. It might be worth taking a look at the 
work we've been doing with OMII-UK (the umbrella organization in the UK 
which has in the last year absorbed both myGrid and OGSA-DAI). They 
(we?) have a container configuration with support for WS-Security 
(certificate based authentication) and will support its use in escience 
projects.

Use of WS 'standards' incurs a cost, both in initial development time 
and in runtime complexity. It potentially reaps a benefit if you go on 
to make use of the full suite (or a substantial subset) of additional 
capabilities such as security, WSRF style session management, full 
message description and the like. Before dropping WS invocation support 
entirely you need to consider the potential future requirements that it 
might fill more easily than a home grown implementation. There is of 
course always the political aspect - we've already had people say "we 
can't use taverna as, although it works and does what we want, it isn't 
a 'standard'". Sad but true.

This isn't an exclusive choice of course, you could (and maybe should) 
have a simple REST-like invocation interface alongside a more complex 
and potentially extensible SOAP based one.

Just my thoughts,

Tom