[moby] Re: [MOBY-dev] Notes from working group 'B'

Martin Senger senger at ebi.ac.uk
Wed Jun 8 15:55:03 UTC 2005


> At the moment, and under the proposed API, there is no defined way for a
> client to interact with the same (replicated) service when it calls an
> async service as it does when it retrieves the result.
>
   And I think that's okay. The method 2,3,4 in the async invocation must 
be treated by clients as one "atomic" operation. If any of them fails, 
client can start again using a new (replicated) URL.
   It, of course, does not preclude some low-level-grid-enabled 
implementations that shared resources between replicated services - and 
that are able to respond by another service instance to a request 
originally sent to a different service instance. But that would be rather 
an exceptional case I guess.

   BTW, I have still on my dest an action from the meeting to send a
proposal how to do this async invocation (so it was me not the Spanich
group who promised to do it). But it seems now that Mark put already
everything on the table - thanks, Mark :-). Anyway, I will try to 
summarize it once again in one of the following emails.

> multiple URLs for the same (authURI,serviceName) combination, then the
> combination of (authURI,serviceName,URL) will uniquely identify a
> service
>
   This is an interesting point because it opens an issue which is not
fully (as far as I can tell) explained in the Moby API: Does the service
URL fully identify a service? Obviously not - I can see several services
with the same URL - and the server engine "adds" the real service name to
this URL. But is this a correct behaviour?
   Surely it works for Perl services - but for Java/Axis services (I 
think) we always add service name to the URL. I guess it works with Perl 
because SOAP::Lite takes the service name from the SOAPAction header. But 
I think that it should not...
   Well, whatever is right, it would be worth to discuss it here (or to
point me to an existing explanation in the Moby API) - and then to
explicitly tell in the API (a) what should be in the URL/endpoint, (b)  
what should be in the SOAPAction header, (c) what should be in the URI
(this one is already fully documented in the Moby API).

   Regards,
   Martin

-- 
Martin Senger

EMBL Outstation - Hinxton                Senger at EBI.ac.uk     
European Bioinformatics Institute        Phone: (+44) 1223 494636      
Wellcome Trust Genome Campus             (Switchboard:     494444)
Hinxton                                  Fax  : (+44) 1223 494468
Cambridge CB10 1SD
United Kingdom                           http://industry.ebi.ac.uk/~senger






More information about the MOBY-dev mailing list