[
https://issues.apache.org/jira/browse/ATLAS-1768?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16025180#comment-16025180
]
Mandy Chessell edited comment on ATLAS-1768 at 5/25/17 6:37 PM:
----------------------------------------------------------------
These are the comments added by [~davidrad] in Jira ATLAS-1410.
210 I wonder if language should be a code table value - or more generally an
valid value from reference data
===> It is a descriptive field that we are not going to process - it is
for human consumption so no value in creating a code table that will need to be
synchronized between all of the metadata repositories. We can add language
identifier - such as "En" is we need to process this in the future.
210 I am wondering about usage. Should this also be a code table - it seems
more structural than the description
==> the usage is a description and examples of how the term is used - it is
a string. Look at dictionary definitions
220 I suggest the supercategory to the subcategory be a composition (filled in
diamond) relationship.
==> This would not be correct because the subcategory is not deleted when
the supercategory is deleted. It remains linked to the Glossary object. I
have added an aggregation (open diamond) to show that the category is
collecting subcategories.
230 I think the GlossaryCategory role name should be categories rather than
category
==> done
240 I wonder about the "to" and "from" ends of the related term as they imply a
direction - for a SYNONYM and TRANSLATION there is no direction. It is almost
like synonyms and transactions should be in a synonym group or translation
group respectively. Maybe we introduce an equivalence group concept, where
everything in the group is related to everything else in the group. This would
help for tag propagation for these terms.
==> I think this over-complicates the model and would make it difficult to
map to IGC. Typically the synonyms are in different glossaries, connecting the
canonical model. I would be nervous about doing tag propagation along
glossary relationships that are not from the spine model.
I don't think we have a way in the current Atlas model to constrain the number
of classifications to 0..1.
==> Classifications have a cardinality - are you saying it does not work? Or
something else?
was (Author: mandy_chessell):
These are the comments added by [~davidrad] in Jira ATLAS-1410.
210 I wonder if language should be a code table value - or more generally an
valid value from reference data
===> It is a descriptive field that we are not going to process - it is
for human consumption so no value in creating a code table that will need to be
synchronized between all of the metadata repositories. We can add language
identifier - such as "En" is we need to process this in the future.
210 I am wondering about usage. Should this also be a code table - it seems
more structural than the description
==> the usage is a description and examples of how the term is used - it is
a string. Look
220 I suggest the supercategory to the subcategory be a composition (filled in
diamond) relationship.
==> This would not be correct because the subcategory is not deleted when
the supercategory is deleted. It remains linked to the Glossary object. I
have added an aggregation (open diamond) to show that the category is
collecting subcategories.
230 I think the GlossaryCategory role name should be categories rather than
category
==> done
240 I wonder about the "to" and "from" ends of the related term as they imply a
direction - for a SYNONYM and TRANSLATION there is no direction. It is almost
like synonyms and transactions should be in a synonym group or translation
group respectively. Maybe we introduce an equivalence group concept, where
everything in the group is related to everything else in the group. This would
help for tag propagation for these terms.
==> I think this over-complicates the model and would make it difficult to
map to IGC. Typically the synonyms are in different glossaries, connecting the
canonical model. I would be nervous about doing tag propagation along
glossary relationships that are not from the spine model.
I don't think we have a way in the current Atlas model to constrain the number
of classifications to 0..1.
==> Classifications have a cardinality - are you saying it does not work? Or
something else?
> Create common types for Open Metadata
> -------------------------------------
>
> Key: ATLAS-1768
> URL: https://issues.apache.org/jira/browse/ATLAS-1768
> Project: Atlas
> Issue Type: New Feature
> Components: atlas-core
> Affects Versions: 0.9-incubating
> Reporter: Mandy Chessell
> Assignee: Mandy Chessell
> Labels: VirtualDataConnector
>
> This JIRA describes a proposal for standard types for open metadata entities
> and relationships. For example, glossaries, database definitions, rules,
> policies, ...
> The value of having standard definitions for metadata is to enable type safe
> APIs and business level UIs plus be able to exchange metadata between
> different instances of metadata repositories.
> The implementation of these common types is divided into 8 areas:
> * Area 0 - for extensions to Apache Atlas's base model
> * Area 1 - for definitions of the data-related assets we are governing and
> using
> * Area 2 - for a glossary of meanings and semantic relationships
> * Area 3 - for information about asset use, crowd-sourced definitions and
> collaboration around the data-related assets
> * Area 4 - for governance such as policies, rules and classifications
> * Area 5 - for reference models and reference data
> * Area 6 - for metadata discovery processes (see
> https://issues.apache.org/jira/browse/ATLAS-1748)
> * Area 7 - for lineage
> Adaptation and flexibility are key in metadata environments so these common
> definitions must be extensible - and we still need to support the ad hoc
> definition of new types in Atlas.
> Apache Atlas supports meta-types that are used in the definition of new
> types. These are currently enumeration, struct, classification and entity.
> JIRA https://issues.apache.org/jira/browse/ATLAS-1690 adds relationships to
> this list. The open metadata models make use of all of these meta-types.
> These are represented by sterotypes on the classes of the open metadata
> definitions.
> The Atlas wiki has the models as a set of linked pages which are probably the
> easiest way to view the models.
> Start here:
> https://cwiki.apache.org/confluence/display/ATLAS/Building+out+the+Apache+Atlas+Typesystem
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
