Meeting in Göteborg 3 October 2013

Time constraints

• Arrival
  • Olha, Przemek and Twan will arrive on Wednesday evening (scheduled arrival 18:05 at Landvetter)
  • Daan will arrive on Thursday around 10:00 am
  • We can start our meeting topics at 9:00am, beginning with technical topics that don't strictly require/interest Daan
• Departure
  • Olha, Przemek and Twan have to be at the airport at 17:55, meaning they will have to leave by bus around 17:00 (according to the information from ​Flygbussarna). The bus stop at Korsvägen is a ten minutes walk from Språkbanken, see ​map.
  • We should end our meeting around 16:30 latest
• Coffee breaks
  • ...

General Agenda

• Hellos

• Short progress report from TLA

• Short progress report from UGOT

• Back-end + client communication status and next steps

• Lunch?

• Web Annotator meeting

• AAI and Sprakbanken meeting

Web annotator sub-agenda

Annotator plugin + Wired Marker conflict

LREC conference paper

Backend issues

• Annotation Body. What do we want there? TLA's proposal (and current implementation): Text attributed by a mimeType.

• At the moment if a resource, e.g. a source, is to be added then it is just added by dao without checking if its content is indeed fresh. I have a simple filter only for users: this filter checks is a user with a given e-mail already exists in the DB. If yes, then it is not added. What could be the criteria of "freshness" for the other resources: CachedRepresentations, Sources, Annotations?

• How would we upload (PUT) and download (GET) blobs of cached representations? E.g. for uploading (POST) we agreed on multipart body. However we have not discussed the schema for such body. So I 'm not implementing the POST method for now. Moreover I'm not sure at all about my GET implementation. The things I have found on the internet so far were not that helpful, it was more for JSP.

• We have an action ADD_CACHED_REPRESENTATION which is a part of a response envelope after adding an annotation. This envelope seems to be not that informative, since it is not clear for which of sources there is no any cached representation (recall that annotations can have multiple sources). I have implemented a dao-method that returns a list of sources without any cached for a given annotation id. However for now it is underused by the REST method, which returns ADD_CACHED_REPRESENTATION is this list is not empty.

• User is a special resource, so I'm not implementing its deletion in Dao. Or??

• Bytea (Postgress) vs blob (HSQL). Are there any other equivalents of blob in Postgress? What would be the better choice? What do we know about bytea?

• In e.g. "GET sources/<sid>/cached/<cid>/metadata" the part source/<sid> is redundant since you can get a cached representation just by its id. However, in the presence of <sid> we would make an additional check if <cid> is indeed one of the cached of <sid>. What would we prefer, a simpler version "GET cached/<cid>/metadata" or more advanced one"GET sources/<sid>/cached/<cid>/metadata" with the check?