[
https://issues.apache.org/jira/browse/OODT-72?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12969200#action_12969200
]
Chris A. Mattmann commented on OODT-72:
---------------------------------------
I fundamentally disagree with this approach. Metadata passed in to Versioners
*is intended* to be immutable.
That's been the architecture as documented since 2005 and there is little
reason to change it. Now onto addressing your specific points:
bq. This prevents Metadata elements such as FileLocation from being set based
off Product References since metadata is added to catalog pre-versioning
No it doesn't. The Versioner can be called ahead of time to compute the file
reference(s) and then flowed through either on client side metadata or server
side.
bq. * A known work-around is to run the versioner twice (once in a
MetExtractor and again when the Versioner is run in the ingest process) . . .
however this is very much a work-around and setting elements such as
FileLocation is a very common functionality.
What you call a work-around I call the intended behavior. Versioners are
intended to get an un-modifiable metadata object. IOW, they shouldn't assume
that any changes they make to the met object will have *any* chance of being
persisted. Period. There was a lot of thinking in this. Here's a
counter-example:
- I create a Versioner to generate a Metadata field that I don't intend to be
persisted, but that I intend to use by sub-classing the
MetadataBasedFileVersioner. This Versioner has a filePathSpec:
{code}
/[ProdDate]/foo/bar/Filename
{code}
where the [ProdDate] met field is the dynamic met field that I intend to
compute in the versioner (but that I don't intend to be persisted).
On the one hand you could say this is dealt with by smart product type policy
management where we simply don't include ProdDate as a metadata element that we
intend to persist, but I'll argue that the met element ProdDate should *not*
even be present in the metadata that's sent to a Catalog to be persisted.
bq. I purpose that Versioning take place before MetExtractors run.
I disagree, -1.
> Unable to set Metadata based off Product Versioning during Product ingestion
> ----------------------------------------------------------------------------
>
> Key: OODT-72
> URL: https://issues.apache.org/jira/browse/OODT-72
> Project: OODT
> Issue Type: Improvement
> Components: file manager
> Affects Versions: 0.1-incubating
> Environment: none
> Reporter: Brian Foster
> Assignee: Brian Foster
> Priority: Minor
> Fix For: 0.2
>
>
> Ingest order:
> - add product to catalog
> - metextractors run
> - add metadata to catalog
> - versioner runs (on client or on server)
> - add references to catalog
> This prevents Metadata elements such as FileLocation from being set based off
> Product References since metadata is added to catalog pre-versioning
> - A known work-around is to run the versioner twice (once in a MetExtractor
> and again when the Versioner is run in the ingest process) . . . however this
> is very much a work-around and setting elements such as FileLocation is a
> very common functionality.
> I purpose that Versioning take place before MetExtractors run.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
