July 2013 Documentation Telecon

July 2013 Documentation Telecon

Attendees: Ted Haberman, Ed Armstrong, Curt Tilmes, Kelly Monteleone, Erin Robinson, Mike McCann, Bob Simon, Phil Jones, Lindsay, Walt Baskin

·         Documentation had sessions at the ESIP summer meeting last week – they went well

o   1st – lessons learned included GRIST, G6, CF & Data issues

o   2nd – ACDD – need to get Dave’s presentation up as there were technical problems on webex.

§  Ted and Ed presented real world examples

o   Dave’s presentation suggested that documentation might move towards CF, but Ed though we had agreed to move to OGC standards

·         ACDD – needs steering committee meeting to discuss name changes (suggested by John)

o   Minimized changes in names in ACDD 1.2 and address these in 2.0

·         Bob – stresses the need for backwards compatibility

o   Avoid duplicating or replacing names

·         Mike – focus on clarity/meaning of terms (need to work on definitions) because people come at this terms from different directions

o   Ex. contributor is principle investigator (PI) in ISO

o   Need better fit in ACDD for people

·         Bob – ‘contributor role’ is gone from website

o   Ted – John added a “information” field and removed ‘contributor role’

·         Mike – want to flatten into ACDD

o   Gulf of Mexico Research group has a tool for ISO metadata that has 3 ‘roles’

o   Ted – there are a large number of roles in ISO, ~20

o   In ACDD need 20 x 3 attribute names (name, email, url)

o   Ed – groups will help organize ‘role’ into discrete ‘objects’

o   ACDD 2.0 will have objects

·         Mike – what about legacy data like NetCDF3

o   If flatten, all people can be put into ‘ResponsibleParty_role_name’ (or email, organization, etc).

o   To have organization in NetCDF3 – need all roles clarified

§  If the attribute is PI_name then parser won’t know that this is a person

§  Also a problem if there are more than 1 PI or user support person

§  HDF has unique names

·         Walt – attribute values in ACDD of a name can be an array of strings

§  Ted – then have to match an array of strings across attributes (… email, position name, institution) – properties of people

§  (primary is 1st in list)

o   Ted – if information designed with delimiters in items – it causes problems with the design (Ed – it gets messy fast)

·         Mike – there are real limitations in flat data models – want a simple flat model

o   New HDF conversion – when group node – can have multiple instances in group with same name

o   ‘groupname_attribute’

·         Ted – maybe ‘_’ is a problem, could use “.” – Mike or “__”…. These all cause other issues

·         When is OceanSite going to use NetCDF4 (asked to Mike)?

o   It isn’t even on the radar – looking at NetCDF3

o   Will have to live with a flat data model

o   Strategy is to use a dumbed down ACDD that might spur a move to NetCDF4

·         Bob – does this mean we need to either flatten ACDD or not support other file types – because groups are only in HDF5 or NetCDF4 extended. 

o   do we need a list of other file types?

·         ACDD 1.2 will be a flat model

·         ACDD 2.0 will have objects (there will be a separate ACDD 2.0 that is flattened)

·         Walt – if 1st had list of group (like a group parameter) that listed prefixes – then could clearly see different between groups and attributes

o   Even if 100s of groups – would be able to decode the structure

o   Like a map of how groups are put together

·         Ted – in terms of flattening mechanisms – we need 3 – 4 to start testing (hopefully not 40-50)

o   Bob – need only 1 flattening method

o   Ted – yes, but need 4 or 5 so we can find the best

·         Mike – need to ensure “lineage” information is not lost when move

o   Ted’s example last week had 9+ objects … all of them are needed

o   xml and Json have methods to parse

o   currently only have object (which is “group”), but need type for group – such as role, lineage, identities. 

·         Bob – has 3 issues

o   Want statement of “backwards compatibility” from this group

o   Limit versions of ACDD

o   Are we really just moving to HDF5 or will ACDD remain for file that do not include groups

·         Mike – felt that ACDD 1.2 is gong backwards

·         Ted – only Ted has commented on John’s discussion and ACDD edits

o   Bob – we need a system to review edits like CF so we can propose changes

o   Such as minicode changes are not added until formal agreement is reached… something like Jira (sp?)

o   Link to ACDD discussion on wiki - http://wiki.esipfed.org/index.php/Talk:Attribute_Convention_for_Data_Discovery_(ACDD)_Working

·         On the ACDD page – the definitions say “an include any information as ISO 19139 or free text”.  This is not helpful, need more information

·         Currently the ‘watch” function on the wiki is not working… but recommend watching the ACDD discussion

·         Next meeting

o   Rich

o   Maybe discussion – ACDD 1.2 finalize and definitions

o   Algorithums for flattening – may need wiki page

o   Other file formats – think about if we don’t need formal discussion of this

·         Next meeting Aug 21 – noon EDT