[Biojava-l] To Do Question

David Block dblock@gnf.org
Tue, 28 May 2002 16:06:00 -0700


UDDI?

Check out http://www.biomoby.org for a cross-platform service registry-in-the-making - not just vaporware, Mark W. has funding for this, and there are some heavyweights in on it.

I would tend to stay in WSDL/UDDI space for this naming stuff, since it integrates with JNDI (right?) and so you don't lose anything, and you get all the SOAP services for free.

--
David Block                    dblock@gnf.org
GNF - San Diego, CA        http://www.gnf.org     
Genome Informatics  /  Enterprise Programming

> -----Original Message-----
> From: Matthew Pocock [mailto:matthew_pocock@yahoo.co.uk]
> Sent: Tuesday, May 28, 2002 3:24 PM
> To: Dickson, Mike
> Cc: 'Chaganthi, Madhusudan R.'; 'smh1008@cus.cam.ac.uk';
> biojava-l@biojava.org; Thomas Down
> Subject: Re: [Biojava-l] To Do Question
> 
> 
> Hi.
> 
> This todo may be partly my fault. It was flagged before JNDI 
> was mature, 
> and it was discussed during an early boot camp. JNDI rocks. URNs/URIs 
> are great too. There are several uses of resolvable names 
> that would be 
> usefull:
> 
> well-known bioinformatics resources:
>    e.g. sequence databases like embl
>    e.g. sequences - perhaps compound id like db.version:seqid.version
> 
> biojava objects
>    e.g. alphabets by name
>    e.g. a sequence you have named
>    e.g. a named HMM
>    e.g. a service like running blast
> 
> Idealy, given a name, BioJava should be able to resolve that to an 
> instance that behaves apropreately. So, the URI for adenosine should 
> resolve to the apropreate symbol object, and the URI for 
> genbank should 
> resolve to an apropreate SequenceDB instance that can be used 
> to fetch 
> genbank records. If you create a novel alphabet (e.g. coin 
> tosses) then 
> another machine on the nework given a URI for this should be able to 
> resolve or build (via RMI, CORBA, SOAP, serialization, whatever) an 
> apropreate alphabet.
> 
> If you have further use cases, please post them.
> 
> The plumbing necisary would be some biojava register allowing 
> names to 
> be associated with objects or factories (and potentialy dropping 
> unreachable objects), and some p2p or RMI or Naming-Directory 
> solution 
> for sharing this infrastructure between VMs that can communicate 
> (presumably via network).
> 
> Should the API or network protocols be open enough to let 
> other Java or 
> other non-Java apps plug in?
> 
> Do we serve up stuff that gets resolved to remote objects, or is it 
> replicated into the name-space of the discovering process?
> 
> Does anyone have vast experience in doing this sort of thing? 
> Has it all 
> been done before?
> 
> Should this be integrated with JavaSpaces or BeanContexts or 
> any other 
> well-known Java APIs?
> 
> Matthew
> 
> Dickson, Mike wrote:
> > I'd guess a few use cases would help.  
> > 
> > What sort of things are we looking up in a directory?  How 
> are names formed?
> > Is there a structure to them, like URN's?  BioJava doesn't 
> really (as far as
> > I'm aware) have a bias towards client or server code.  
> Where in the codebase
> > would we expect directory and naming to get used?
> > 
> > I don't have the answers, these are just questions I'd be 
> inclined to ask.
> > 
> > Mike
> > 
> > 
> >>-----Original Message-----
> >>From: Chaganthi, Madhusudan R. [mailto:mvc9@cdc.gov]
> >>Sent: Tuesday, May 28, 2002 4:55 PM
> >>To: 'Dickson, Mike'; Chaganthi, Madhusudan R.; 
> 'smh1008@cus.cam.ac.uk'
> >>Cc: biojava-l@biojava.org; Thomas Down; Matthew Pocock
> >>Subject: RE: [Biojava-l] To Do Question
> >>
> >>I personally feel that JNDI is the way to go. I will be 
> interested in
> >>hearing other suggestions too.
> >>
> >>My question is, has there been a vision about the hows and 
> whats for this
> >>feature?
> >>
> >>Its too broad a feature for someone to envision. Maybe we 
> need to break it
> >>up into chewable pieces.
> >>
> >>What do you say?
> >>
> >>
> >>-----Original Message-----
> >>From: Dickson, Mike [mailto:mdickson@netgenics.com]
> >>Sent: Tuesday, May 28, 2002 4:58 PM
> >>To: 'Chaganthi, Madhusudan R.'; 'smh1008@cus.cam.ac.uk'
> >>Cc: biojava-l@biojava.org; Thomas Down; Matthew Pocock
> >>Subject: RE: [Biojava-l] To Do Question
> >>
> >>
> >>Is there a reason we wouldn't just use JNDI for this?
> >>
> >>
> >>>-----Original Message-----
> >>>From: Chaganthi, Madhusudan R. [mailto:mvc9@cdc.gov]
> >>>Sent: Tuesday, May 28, 2002 4:10 PM
> >>>To: 'smh1008@cus.cam.ac.uk'; Chaganthi, Madhusudan R.
> >>>Cc: biojava-l@biojava.org; Thomas Down; Matthew Pocock
> >>>Subject: RE: [Biojava-l] To Do Question
> >>>
> >>>Ok.
> >>>
> >>>If I can get any formal requirements documentation on 
> this, that will be
> >>>helpful for me.
> >>>
> >>>
> >>>
> >>>-----Original Message-----
> >>>From: David Huen [mailto:david.huen@ntlworld.com]
> >>>Sent: Tuesday, May 28, 2002 3:28 PM
> >>>To: Chaganthi, Madhusudan R.
> >>>Cc: biojava-l@biojava.org; Thomas Down; Matthew Pocock
> >>>Subject: Re: [Biojava-l] To Do Question
> >>>
> >>>
> >>>On Tuesday 28 May 2002 7:37 pm, Chaganthi, Madhusudan R. wrote:
> >>>
> >>>>Longer term features
> >>>>*	Naming and directory services throughout
> >>>>
> >>>>What exactly are the features that are to be supported 
> here in this
> >>>>context? If I find the time at home, I will be more than 
> happy to help
> >>>>you out in this area.
> >>>>
> >>>>Thanks
> >>>>Madhu
> >>>>
> >>>
> >>>I don't think we have any yet although both Matt and Thomas have
> >>
> >>discussed
> >>
> >>>it.  Is there even an agreed standard amongst the Bio* projects for
> >>
> >>these
> >>
> >>>services yet?
> >>>
> >>>Regards,
> >>>David
> >>>_______________________________________________
> >>>Biojava-l mailing list  -  Biojava-l@biojava.org
> >>>http://biojava.org/mailman/listinfo/biojava-l
> >>
> > 
> 
> 
> 
> _______________________________________________
> Biojava-l mailing list  -  Biojava-l@biojava.org
> http://biojava.org/mailman/listinfo/biojava-l
>