- The very first step is actually to define the core entities.

- The relation of the Entities must be defined in the infomodel, furthermore the different implementations (ldap, xml, sql) should come with a properly defined structure which follows (more or less) the abstract relation of the entities. Entities should be reflected in the implementation, for example entities can correspond to ldap objectclasses. The implementation (ldap schema & DIT, xml schema, etc..) themself should be part of the 'standard'

- The computing service should include the description of the hardware behind (or be associated to). One of the biggest problem of glue-1.x is the fact that the CE is decoupled from the actual hardware.

- No need to expose internal computing service (cluster) properties such as cluster.tmpdir, cluster.wntmpdir, software pathes

- Authentication info should be present (certificates, CAs) in the service description

- The schema should avoid using middleware specific attributes (e.g. Gram version, gatekeeper port, etc..)

- The way the authorization is presented should be consistent for every service, try to avoid introducing CEAuthorization, SEAuthorization, BrokerServiceAuthorization. No need for Ce.Policy and SE.Policy, let's use a common 'policy'

- The way resource characteristics are presented in the schema should relate to the job description, Infoschema and job description should be synchronized. An example job memory requirement -> advertised resource memory JSDL should be studied

- The grid job entity at least should standardize the job states

- naming: avoid the usage of unnecessarily long empty often confusing strings: StateStatus, InfoContactString

- take care of the attribute semantics: "UniqueID = string" is rather ambiguous

- last but not least, the new model should be powerful enough to accommodate all the current info present in the ARC schema. Quick list of ARC objects and attributes (ignoring the storage element and replica catalogue objects):

• Object nordugrid-cluster: nordugrid-cluster-name

• Object nordugrid-queue: nordugrid-queue-name nordugrid-queue-status nordugrid-queue-running nordugrid-queue-queued nordugrid-queue-maxrunning

• Object nordugrid-job: nordugrid-job-globalid nordugrid-job-globalowner nordugrid-job-status nordugrid-job-queuerank nordugrid-job-submissionui nordugrid-job-submissiontime nordugrid-job-usedcputime

• Object nordugrid-authuser: nordugrid-authuser-name nordugrid-authuser-sn nordugrid-authuser-queuelength nordugrid-authuser-diskspace nordugrid-authuser-freecpus