RE: Re: [ogsa-d-wg] Discovery Service - Returning Data
Allen, I'm also a bit concerned about this. I apologise, because I didn't realise when we were discussing this on the call that you had taken the action to make this change - I was in too much of a hurry to move on to the next item. My concern is that I don't know what it means to "return the data", in a context where we have several possible interfaces for accessing data, including ByteIO and the various DAIS specs, each of which operate on names. I don't think I understand Dieter's suggestion sufficiently well to see how it would affect the overall architecture. It certainly sounds an interesting idea and if we can incorporate it simply then that's all to the good, but I'd rather we had a longer talk with Dieter first. Best wishes, Dave. _____ From: owner-ogsa-d-wg@ggf.org [mailto:owner-ogsa-d-wg@ggf.org] On Behalf Of Allen Luniewski Sent: 16 March 2006 16:59 To: ogsa-d-wg@ggf.org Subject: Fw: Re: [ogsa-d-wg] Discovery Service - Returning Data Peter's mail bounced. Allen ----- Forwarded by Allen Luniewski/Almaden/IBM on 03/16/2006 08:54 AM ----- Message-ID: <20570961.451142526276621.OPEN-XCHANGE.WebMail.tomcat@sala.cscs.ch> Date: Thu, 16 Mar 2006 17:24:36 +0100 (CET) From: Peter Kunszt <pkunszt@cscs.ch> To: ogsa-d-wg@ggf.org Subject: Re: [ogsa-d-wg] Discovery Service - Returning Data hi allen i don't know.. i get an uneasy feeling when service discovery is mixed with ontologies and content. to me, e.g. DNS is service discovery already, and certainly it does not give me the semantics or content of the service. i would separate any kind of content retrieval from service discovery, it feels 'cleaner' for a SOA. otherwise i'm afraid service discovery will be too large a beast to manage.. just my swiss 2cents... cheers peter -----Urspr=C3=BCngliche Nachricht----- Von: owner-ogsa-d-wg@ggf.org im Auftrag von Allen Luniewski Gesendet: Mi 15.03.2006 18:26 An: ogsa-d-wg@ggf.org Betreff: [ogsa-d-wg] Discovery Service - Returning Data =C2=A0 At today's call, I took on an action to update the discovery service section to allow for discovery services that returned the data of interest rather than a reference to the data.=C2=A0 Here is the updated material for your=C2=A0 consideration: Resource Discovery Discovery services are vital to the data architecture.=C2=A0 The data services may use the discovery services not just for registering services themselves, but also for registering the data sets that are stored by those services. This requires languages or ontologies for describing data. Discovery services may also register the locations of schema definitions. Discovery services may be built upon metadata repositories that contain information about other entities in the grid such as resources, and services.=C2=A0 Some discovery services may return the name of the service and some may return an EPR to the service.=C2=A0 Some discovery services may ta= ke a description of the desired data as input and return the data itself. Allen
Hi, Is it not the case that you may want some higher level service that given some metadata (or attributes) about the type of data that you are looking for will go and talk to services (discovery and data services), locate the data that matches your criteria and return it to the user. The user never be concerned about what the physical or logical name for that data was. In effect some sort of user broker or agent. That was my understanding from the call. I think this is a higher-level service that uses discovery services but is not necessarily a registry. I would agree with Peter that we do not want to bind the discovery process with the content retrieval process in the architecture. On the other hand if someone wants to create such a combined beast then we should not necessarily put obstacles in their way but that would be outwith the scope of the architecture. Mario On Thu, 16 Mar 2006, Dave Berry wrote:
Allen,
I'm also a bit concerned about this. I apologise, because I didn't realise when we were discussing this on the call that you had taken the action to make this change - I was in too much of a hurry to move on to the next item. My concern is that I don't know what it means to "return the data", in a context where we have several possible interfaces for accessing data, including ByteIO and the various DAIS specs, each of which operate on names.
I don't think I understand Dieter's suggestion sufficiently well to see how it would affect the overall architecture. It certainly sounds an interesting idea and if we can incorporate it simply then that's all to the good, but I'd rather we had a longer talk with Dieter first.
Best wishes,
Dave.
_____
From: owner-ogsa-d-wg@ggf.org [mailto:owner-ogsa-d-wg@ggf.org] On Behalf Of Allen Luniewski Sent: 16 March 2006 16:59 To: ogsa-d-wg@ggf.org Subject: Fw: Re: [ogsa-d-wg] Discovery Service - Returning Data
Peter's mail bounced.
Allen
----- Forwarded by Allen Luniewski/Almaden/IBM on 03/16/2006 08:54 AM -----
Message-ID: <20570961.451142526276621.OPEN-XCHANGE.WebMail.tomcat@sala.cscs.ch> Date: Thu, 16 Mar 2006 17:24:36 +0100 (CET) From: Peter Kunszt <pkunszt@cscs.ch> To: ogsa-d-wg@ggf.org Subject: Re: [ogsa-d-wg] Discovery Service - Returning Data
hi allen
i don't know.. i get an uneasy feeling when service discovery is mixed with ontologies and content. to me, e.g. DNS is service discovery already, and certainly it does not give me the semantics or content of the service. i would separate any kind of content retrieval from service discovery, it feels 'cleaner' for a SOA. otherwise i'm afraid service discovery will be too large a beast to manage..
just my swiss 2cents...
cheers peter
-----Urspr=C3=BCngliche Nachricht----- Von: owner-ogsa-d-wg@ggf.org im Auftrag von Allen Luniewski Gesendet: Mi 15.03.2006 18:26 An: ogsa-d-wg@ggf.org Betreff: [ogsa-d-wg] Discovery Service - Returning Data =C2=A0 At today's call, I took on an action to update the discovery service section to allow for discovery services that returned the data of interest rather than a reference to the data.=C2=A0 Here is the updated material for your=C2=A0 consideration: Resource Discovery Discovery services are vital to the data architecture.=C2=A0 The data services may use the discovery services not just for registering services themselves, but also for registering the data sets that are stored by those services. This requires languages or ontologies for describing data. Discovery services may also register the locations of schema definitions. Discovery services may be built upon metadata repositories that contain information about other entities in the grid such as resources, and services.=C2=A0 Some discovery services may return the name of the service and some may return an EPR to the service.=C2=A0 Some discovery services may ta= ke a description of the desired data as input and return the data itself.
Allen
+-----------------------------------------------------------------------+ |Mario Antonioletti:EPCC,JCMB,The King's Buildings,Edinburgh EH9 3JZ. | |Tel:0131 650 5141|mario@epcc.ed.ac.uk|http://www.epcc.ed.ac.uk/~mario/ | +-----------------------------------------------------------------------+
Dave, No problem - I guess that I misunderstood. Obviously easy enough to take out the "return data" text. I thought that the suggestion to return data was reasonable even though I doubt that this will be a common thing to do. Allen "Dave Berry" <daveb@nesc.ac.uk> Sent by: owner-ogsa-d-wg@ggf.org 03/16/2006 09:30 AM To <ogsa-d-wg@gridforum.org> cc Subject RE: Re: [ogsa-d-wg] Discovery Service - Returning Data Allen, I'm also a bit concerned about this. I apologise, because I didn't realise when we were discussing this on the call that you had taken the action to make this change - I was in too much of a hurry to move on to the next item. My concern is that I don't know what it means to "return the data", in a context where we have several possible interfaces for accessing data, including ByteIO and the various DAIS specs, each of which operate on names. I don't think I understand Dieter's suggestion sufficiently well to see how it would affect the overall architecture. It certainly sounds an interesting idea and if we can incorporate it simply then that's all to the good, but I'd rather we had a longer talk with Dieter first. Best wishes, Dave. From: owner-ogsa-d-wg@ggf.org [mailto:owner-ogsa-d-wg@ggf.org] On Behalf Of Allen Luniewski Sent: 16 March 2006 16:59 To: ogsa-d-wg@ggf.org Subject: Fw: Re: [ogsa-d-wg] Discovery Service - Returning Data Peter's mail bounced. Allen ----- Forwarded by Allen Luniewski/Almaden/IBM on 03/16/2006 08:54 AM ----- Message-ID: <20570961.451142526276621.OPEN-XCHANGE.WebMail.tomcat@sala.cscs.ch> Date: Thu, 16 Mar 2006 17:24:36 +0100 (CET) From: Peter Kunszt <pkunszt@cscs.ch> To: ogsa-d-wg@ggf.org Subject: Re: [ogsa-d-wg] Discovery Service - Returning Data hi allen i don't know.. i get an uneasy feeling when service discovery is mixed with ontologies and content. to me, e.g. DNS is service discovery already, and certainly it does not give me the semantics or content of the service. i would separate any kind of content retrieval from service discovery, it feels 'cleaner' for a SOA. otherwise i'm afraid service discovery will be too large a beast to manage.. just my swiss 2cents... cheers peter -----Urspr=C3=BCngliche Nachricht----- Von: owner-ogsa-d-wg@ggf.org im Auftrag von Allen Luniewski Gesendet: Mi 15.03.2006 18:26 An: ogsa-d-wg@ggf.org Betreff: [ogsa-d-wg] Discovery Service - Returning Data =C2=A0 At today's call, I took on an action to update the discovery service section to allow for discovery services that returned the data of interest rather than a reference to the data.=C2=A0 Here is the updated material for your=C2=A0 consideration: Resource Discovery Discovery services are vital to the data architecture.=C2=A0 The data services may use the discovery services not just for registering services themselves, but also for registering the data sets that are stored by those services. This requires languages or ontologies for describing data. Discovery services may also register the locations of schema definitions. Discovery services may be built upon metadata repositories that contain information about other entities in the grid such as resources, and services.=C2=A0 Some discovery services may return the name of the service and some may return an EPR to the service.=C2=A0 Some discovery services may ta= ke a description of the desired data as input and return the data itself. Allen
participants (3)
-
Allen Luniewski -
Dave Berry -
Mario Antonioletti