[
https://issues.apache.org/jira/browse/ATLAS-1690?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15967442#comment-15967442
]
David Radley commented on ATLAS-1690:
-------------------------------------
Thank you very much [~Stefhan] for the quick thorough feedback. REsoponding to
you feedback in <<david >> tags :
Page 1:
• “Metadata stores store metadata” re-write this to “Metadata
repository stores metadata” as it is less confusing. <<David agreed>>
• Suggestion to replace ‘metadata entity’ with ‘metadata
object’, as the term object is more generic. <<David agreed>>
• “Relationships include when Glossary Terms are linked to each
other, assets mapped to Glossary Terms and asset metadata linked together.” Are
there types of relationships between metadata objects that are excluded from
this scope? <<David Category to term, glossary to term, as we bring in
classification schemes - there will be more relationships. I just wanted to
give a flavor of the the relationships ) >>
Page 2:
• “As a 1 When this document refers to Glossary Term – it is
referring to those described in Jira 1410. Page 2 of 9 Glossary Term is linked
to an asset, we need to be aware that there are 3 owners, one for the Glossary
Term one for the asset and one for the relationship. This document proposes a
design that allows relationships to be considered in this way.” Does this state
that we consider relationships as ‘first class citizens’? As such that the
relationship between two objects becomes an object on its own, with a unique
identifier and potentially additional metadata? <<David yes for the loose
relationship case>>
• “When this document refers to Glossary Term – it is referring
to those described in Jira 1410.” This exact sentence was already mentioned as
footnote on page 1. <<David good spot I will remove >>
Page 5:
- “Relationship constraints can have an attribute name and a
reverse attribute name, but there is no label for the relationship itself – the
relationship name.” What about synonym relationships? <<David I am thinking of
handing these semantic relationships in the glossary document. My thoughts at
the moment are that we would model all semantic relationships as bidirectional
relationships in the type system. For synonyms, I wonder if we could have a
many to many relationship specifying the glossary term synonym as both ends of
the relationship - this is a great unit test case for this item! >> What would
then be the attribute name and the reverse attribute name? <<David synonym>>
- With regards to the Bidirectional relationships example. It
would be: ‘A person has a home address’/’A home address is of a person’. The
‘has a’ / ‘is of’ are two directional relationships that would always need to
be kept in sync. <<David So the has-a and is-of relationships are semantic - so
not in this doc. I think we should model semantic relationships as
bidirectional type relationships.so we would knot the 2 ends togetherin ot he
one relationship. >> ‘A home address is a address’. A more accurate example of
a bidirectional relationship would be ‘First name is synonym of Given name’ /
‘Given name is synonym of First name’ <<David I am thinking of these as other
examples of bidirectional relationships>>
- With regards of the styles of relationships, perhaps its good
idea to adopt the terminology of:
o Hierarchical relationships (inheritance)
o Equivalence relationships (synonyms/is a)
o Associative relationships (related to / loosely coupled)
https://twobenches.wordpress.com/2008/06/04/semantic-relationships-in-a-thesaurus/
For inspiration see: Morville, Peter and Rosenfeld, Louis (2002)
http://yunus.hacettepe.edu.tr/~tonta/courses/fall2010/bby607/IAWWW.pdf
Page number 203, page 225, Definition ‘thesaurus’: A controlled vocabulary in
which equivalence, hierarchical, and associative relationships are identified
for purposes of improved retrieval.*
* Guidelines for the Construction, Format, and Management of Monolingual
Thesauri. ANSI/NISO Z39.19–1993 (R1998).
See also Figure 9-12. Semantic relationships in a thesaurus (Page number 204,
page 226).
See also Semantic Relationships (page number 215, page 237)
<<david - I see these as different sorts of semantic relationships. These are
different to semantic classification, relationships around classification
schemes and relationships between structural metadata. This document is
concerned with the base structural metadata relationships - i.e. relationships
the base types system is aware of. This is more of uml class diagram world of
classes, objects and associations; so this document is talking of the uml style
associations - this will be the foundation on which the other relationship
types can be implemented. does this make sense?>>
Page 7:
- Delete functionality. This is crucial as it will allow to cope
with changes in the metadata objects, and relationships provided by other
metadata repositories (e.g. other atlas instances).
- With regards to the discussion point on endpoints. Endpoints
should not be updated / modified when working on relationships. The
relationship should point to the object and subject identifier without
impacting those objects in any way. A separate mechanism should be responsible
for interpreting the relationships and if needed adjusting/updating the
associated objects. E.g. collapsing the assigned classification into the
subject or one of the endpoints of the subject (e.g. assigned information
view). <<david - agreed - we are scoping this document to the base type system
relationships>>.
Page 9:
- On ‘1 to 1 relationships’: I do not understand why this would
be an issue, if you would define the relationship object independently of the
endpoints. In that way, both endpoints need to exist (otherwise, you will not
even be able to select them) before you can create the relationship. Only the
relationship object will then be created pointing to the identifiers of both
the subject and object. <<david but how can you create the 2 objects without
the relationship being in place. To do this you would need to create all 3 at
the same time - as creating any one of them by itself is invalid; or we need to
tolerate an invalid state for a time. Am I missing something? We can to 1 to 1
optional relationships as you say. >>
> Introduce top level relationships
> ---------------------------------
>
> Key: ATLAS-1690
> URL: https://issues.apache.org/jira/browse/ATLAS-1690
> Project: Atlas
> Issue Type: Improvement
> Reporter: David Radley
> Assignee: David Radley
> Labels: VirtualDataConnector
> Attachments: Atlas Relationships proposal v1.0.pdf
>
>
> Introduce top level relationships including support for
> -many to many relationships
> - relationship names including the name for both ends and the relationship.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
