[ 
https://issues.apache.org/jira/browse/ATLAS-2456?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16368955#comment-16368955
 ] 

David Radley commented on ATLAS-2456:
-------------------------------------

[~sarath.ku...@gmail.com] I have not had time to go through the code in a lot 
of detail. I think it looks pretty good. I wanted to check how you handle the 
following situations:

1) If there is a classification that is applied to an entity with certain 
attributes defined, and you decide to propagate the same classification to that 
entity, in this situation there must be 2 classification instances - each with 
different property values. 

So how does the code handle the case when there are 2 instances of the same 
classification with different property values that propagate to the same target 
entity.

2) I think the classifications defined in the open type 
GovernanceActionclassifications 
[https://cwiki.apache.org/confluence/display/ATLAS/Area+4+-+Governance] are 
going to be the key classifications around governance. There classifications 
are defined by their enums- which have an implied low-high order.  So for 
example confifidentialityLevel of sensitive is higher than internal. For point 
1 scenario- we would want the higher value to win.    

4) I think that any classification that is explicitly defined on an entity wins 
over propagated values.

5) Can you confirm that only one classification of a type is allowed for a 
given entity.

6) I assume that the entitytypes defined on a classificationdef will always 
constrain - so classification and propagated classification (or their subtypes) 
can only ever classify entities of those types (or their subtypes). 

7) I assume that a classification and subtypes of a classification could both 
be applied to an entity. This could compromise the idea of one classification 
per entity. How do you see this working or is this policed as invalid?  For 
example a Confidentiality classification and its subclass could be applied to 
an entity. - each specifying different levels of confidentiality - which one is 
effective?

> Implement tag propagation using relationships
> ---------------------------------------------
>
>                 Key: ATLAS-2456
>                 URL: https://issues.apache.org/jira/browse/ATLAS-2456
>             Project: Atlas
>          Issue Type: New Feature
>          Components:  atlas-core
>    Affects Versions: 1.0.0
>            Reporter: Sarath Subramanian
>            Assignee: Sarath Subramanian
>            Priority: Major
>             Fix For: 1.0.0
>
>         Attachments: TAG PROPAGATION IN ATLAS v.1.pdf, 
> tag_propagation_rest_api
>
>
> {color:#172b4d}Scalable way to quickly and efficiently propagate tags for 
> efficient searches and tag based security. Likewise tags for derivative 
> dataset should be inherited from the parent. For example, if an entity is 
> tagged "PII" then resulting entity created from a CTAS operation should also 
> be tagged "secret" to maintain the classification of the parent. In the case 
> where 2 or more datasets are aggregated the derivative dataset should be a 
> union of all parent tags.{color}



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

Reply via email to