The tables on this page summarise decisions taken during first the SNAP, then the SimDB projects. These issues were discussed mainly in a small core group ("tiger team") of developers who on and off included Laurent Bourges, Igor Chillingarian, Claudio Gheller, Norman Gray,Franck LePetit, Mireille Louys, Patrizia Manzato, Rick Wagner, Herve Wozniak (and others who have not been forgotten).

For the further development of SimDB (and possibly SimDAP) it may be useful to list these and possibly reevaluate them. The issues are subdivied in a few subjects:

• definition of SimDB as a DAL service
• data modelling methodology
• data model contents
• SimDB as a TAP service
• SimDB and Registry

Definition of SimDB as a DAL service

(starting with SNAP)

Item	OK	NOT OK	Comment
SNAP is a DAL protocol for simulations1			Initiated in theory workshop in Cambridge Feb 2006
SNAP follows SSA: data model + protocol2			Victoria interop, 2006
SNAP data model complex, queryData requires registry-like approach			Discussed in April 2007 SNAP meeting of core participants (up till then), Garching
SNAP separated into SimDB and SimDAP			formally announced in 2008 Trieste interop, agreed by "tiger team" before
SimDB is a protocol for finding interesting simulations and postprocessing experiments and their results
SimDB implements "queryData" part of "simple" DAL protocols
"getData" part is implemented by SimDAP or custom services
SimDB development uses Volute			Decision "tiger team", early 2008

¹ The idea was to follow SIAP and apply it to N-body simulations. Later expanded to include other 3+1D simulations such as Adaptive Mesh Refinement (AMR).

² First attempt was SIAP based, but by Victoria interop SSA was the preferred (input from DAL group) approach.

SimDB data modelling methodology

Item	OK	NOT OK	Comment
SimDB/DM represented in UML class diagram1			Follows decision in Cambridge interop 2003.
SimDB/UML is self documenting2			This was decision of core designers (GL, RW, LB), presented in 2008 Trieste Interop
SimDB/DM's UML stored as XMI document (ASCII)			XMI is standard XML format for representing UML documents.
SimDB uses MagicDraw Community Edition 12.1 for the XMI44

|!SimDB/DM uses a predefined UML profile^{[[ ][5]]} | | | Simplicity, standard approach to domain specific languages. |

Propose an intermediate XML format for representing DMs following profile.			Originated in VO-URP development. In practice very useful. Very much more readable than XMI. in principle hand-editable. Core of VO-URP TAP/XSD/HTML/UTYPE/Java/DDL/... generation.

¹ Though the decision in Cambridge never went into any detail, we use Class Diagram as that is best able to represent the static structure of data models that need to be represented in different formats (XML, TAP_SCHEMA, etc).

²All entities and classes must be properly documented. SimDB/Note gives background and context, UML model should be sufficient to define complete model.

⁴ Freely available tool, ESO has commercial version. Should not be necessary to specify a tool, as XMI should deal with that, but in practice useful to "insist" on one

⁵Aimed to simplify models and promote interoperability by presenting uniform DM langugage fo the IVOA. Restriction + extension of UML, currently informal representation (Appendix in Note) and formal implementation (a UML2 profile generated by ESO, updated in course of design).

SimDB/DM contents

Item	OK	NOT OK	Comment
SimDB aims at describing "large/3+1D simulations"			In principle only Snapshot iso more generic Result class is explicitly 3+1D
"getData" services can be foud in SimDB

SimDB/TAP

Item	OK	NOT OK	Comment
SimDB is a TAP service			first discussed April 2007 Garching
SimDB's TAP_SCHEMA predefined			Allows for integration: same ADQL can be sent to all SimDB services.
SimDB/TAP_SCHEMA derived from SimDB/DM using predefined OR-mapping rules
SimDB/TAP's OR mapping rules uses

SimDB and Registry

Item	OK	NOT OK	Comment