[MOBY-l] Re: Welcome!

[Catherine Letondal]

Mark Wilkinson wrote:
> Hi Catherine!
> 
> Welcome to the moby-dev list.  If you have not already done so, please
> also register for moby-l, since this is where most common discssion
> takes place.
> 
> Since moby-dev is intended for core MOBY developers, and we often send
> out open invitations to conference calls and so forth, I like to ask
> everyone who subscribes to the -dev list to post a message introducing
> themselves and explaining what types of development they plan to do.  I
> discourage "lurking" on that list :-)
> 
> Best wishes!
> 
> Mark

Hi,

I like very much the BioMoby project, and, apart from participating in
the very interesting discussions that take place in this list,  
a reason why I'm interested in participating in this project is
for the changes I wish to make in the next version of Pise, an
interface generator I have built
(http://www.pasteur.fr/recherche/unites/sis/Pise/).   
In fact I intend to change it from scratch, and one of the issues that
I feel are very important, is to have a system that would
integrate, not only locally installed programs, as does the current
version of Pise, but also various remote analysis services.

I have read BioMoby (and myGrid, ISYS and biopipe and other) papers
documentation and discussions, and it appears that it is a good
framework for this development, for 2 reasons:
 - Pise is based on various types of interface generators (Web, perl
API, Tcl, Python, SeWer,..), and developping a generator for BioMoby-like
services and install them here at Pasteur for the analysis services we
provide, should be feasible,
 - we wish to adopt the Web services approach,
 - the integration of distributed services within a user interface, that 
we wish to provide, would benefit from a service discovery mechanism, and 
the one that is used in BioMoby (with a view on myGrid as far as I have 
understood?) could suits our needs.

But there is something else we would like to develop, and I would be
happy to get some feedback from people in this list about it. We
really would like to make other Web servers available to users of Pise
in the same user interface and at the same level of choice. But one 
difficulty for a server (so I mean any Web server) to be available in
BioMoby is that you need to make a wrapper, install SOAP::Lite,
etc... on a public server (as explained in
http://www.biomoby.org/SettingUpServices.html). This is 
not always feasible and not every analysis service provider will be OK
to take the time to install everything for this to work. 

So one possibility could be to have this BioMoby wrapping performed and
installed independantly, e.g having the wrapping in an independant server
(maybe what L. Stein calls the knuckles-and-nodes approach?), acting as a
kind of proxy for the legacy server;
http://www.pasteur.fr/~letondal/tmp/kn.png illustrates this idea.

One solution could be to have this wrapper generated by Pise: Pise DTD
already describes how to wrap a local standalone program, so there it should
be feasible to describe how to wrap a Web server. This description is
not exactly the same as a WSDL description, since its role is to describe the
correspondances between two interfaces: the generic one, conforming to
BioMoby, and the legacy Web server one. For this purpose, WIDL (Web
Interface Definition Language) might be interesting (it is a note to the W3C
- see http://www.w3.org/TR/NOTE-widl-970922). I don't know whether it
has been adopted, I believe not, but it is similar to the concept that I'm
looking for. 

So, I don't know whether such an idea has already been discussed, or
if someone has already worked on it?

Thanks in advance for your feedback!


--
Catherine Letondal -- Pasteur Institute Computing Center