[ 
http://opencast.jira.com/browse/MH-7415?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=26218#comment-26218
 ] 

Tobias Wunden commented on MH-7415:
-----------------------------------

The fields that we are currently writing to the search index seem to be working 
for everybody. So I suggest that we move forward with it. The goal is not that 
we get a full abstraction for all possible metadata fields in each and every 
catalog format. The goal is simply to

1) Remove the assumption from search that there must be a dublin core catalog 
in the mediapackage, since the metadata might as well livein another catalog 
type. The way to do this is to define an api that is a s simple and 
straight-forward as possible and exposes only access to fields that go into 
thesearch index. We can refine and expand that api anytime.

2) Fix the fact that we are currently only storing one of possibly multiple 
values in certain fields such as the creator. In addition, allow for multiple 
values for the same field coming from different catalog types. If the catalogs 
are contradicting each other, then that's a problem outside of the search 
service and needs to be fixed there.

Regarding the two sets of metadata: the answer is to *not* use the metadata 
from the mediapackage, as it has been compiled from the mediapackage catalogs 
using the mediapackagemetadata services. I know that right now we are doing 
this, but this should be replaced by the discussed metadata extraction service.

As for whether we should be creating generic search services, the answer is 
"no, absolutely not". If people need a different search index, they need to 
create their own search service.

Since this task seems to be ever-growing we should for now drop support for 
multiple languages and add that at a  later point, as it will be challenging to 
expose these capabilities in the search queries. In addition, this task should 
focus on static metadata (a la dublin core) only and not tackle the direct 
references to mpeg-7. Since the end of this iteration is eventually coming, it 
would be great to get an initial proposal for the aforementioned problems in 
place soon.




> Add support for multiple field values in search index
> -----------------------------------------------------
>
>                 Key: MH-7415
>                 URL: http://opencast.jira.com/browse/MH-7415
>             Project: Matterhorn Project
>          Issue Type: Task
>          Components: Architecture & Services
>    Affects Versions: 1.1 
>            Reporter: Tobias Wunden
>            Assignee: Christoph E. Driessen
>             Fix For: 1.2
>
>         Attachments: 0001-made-DCMIPeriod-immutable.patch, 
> bugfix_SolrUtils.patch
>
>
> Right now, only the first value of multiple possible values of dublin core 
> documents make it into the search index. This needs to be fixed by adding all 
> available field values, eventually by turning the respective fields either 
> into indexed fields (title_en, title_de) or multivalue fields (or both).

-- 
This message is automatically generated by JIRA.
-
For more information on JIRA, see: http://www.atlassian.com/software/jira

        
_______________________________________________
Matterhorn mailing list
[email protected]
http://lists.opencastproject.org/mailman/listinfo/matterhorn


To unsubscribe please email
[email protected]
_______________________________________________

Reply via email to