Thurs from 10-3 CDT are free for me.
At 09:35 AM 5/2/2005, Robert E. McGrath wrote:
Russ suggested Thursday this week for a telcon. Is anyone else available?
I have updated the status report I sent Friday. Quincey gave me more info.
on the last two items, group/indexing and UTF-8.
On 2005.04.29 10:29 Robert E. McGrath wrote:
I think we planned to have a telcon pretty soon, to synch up the netcdf4/hdf5
activities. I would think the agenda is mainly updates.
Mike et al. visited NASA last week, and there has been a lot of work done.
I attach below a summary of big items from HDF. (This is my version, I
hope I got things more or less correct.)
Status (also at http://hdf.ncsa.uiuc.edu/RFC/netCDF4/nc4status.txt)
Reviewing my mail for the last few weeks, I have 4 major items from HDF:
1. Draft proposal for configuration/build
2. Parallel test plan
3. Discussions of group and indexing features and plans
4. Discussions of unicode
Here is my summary
Draft proposal: "A Roadmap for Implementing NetCDF4 Configure/build"
(March 30, 2005)
See the document:
Status: I think this document represents the consensus of the project.
At some point, it will be important to publish something.
Parallel test plan and test status
"Ideas for Testing Parallel NetCDF-4" (April 12, 2005, revised April
See the document:
(Note: see the _revised_ plan on page 3-4)
Status: This work is in progress.
Discussion of Group Indexes
* For feature freeze: implement new API with old index internally *
Reimplement with new structures for the release
The files created in this period may be incompatible with the released
Multiple options for what will stored for each link (i.e., the index
in a group):
- Name (indexable, must be unique, modifiable)
- Creation time (indexable, may be non-unique, modifiable)
- Creation order (indexable, unique, monotonicly increasing,
- Character set (i.e. ASCII, UTF-8, etc.) (non-indexable?,
- Object address/link target (for hard/soft links) (non-indexable, may be
non-unique (for multiple links to same object), modifiable?)
Applications can determine which of the three indexable fields they'd like
to have an index maintained for with a group creation property.
choose an index for iteration, etc. with a group access property.
Additional requirements have been proposed:
* No current API will change, all existing programs work without change.
* No matter what index is selected at create time it must be possible to
iterate by name, and this must be the default behavior.
I gather that these requirements may not be fully met, but no details
are available at this time.
Discussions of Unicode
Unidata requested support for Unicode, at least for object (and
attribute?) names. This was limited to UTF-8 support for now.
An RFC is being prepared Summary of approach:
HDF should be able to accept UTF-8 encoded file names, object names,
attribute names and strings as a datatype. We are running tests
to verify this.
Add several small features to the library to allow the file
to have better self-description information about UTF-8 strings:
- A field in the link to an object that describes the character
set encoding used for the name
- Some metadata on attributes describing the character set
encoding used for the name
- A new character set encoding type for string datatypes.
Details will be presented in an RFC, RSN
Mike Folk, Scientific Data Tech (HDF) http://hdf.ncsa.uiuc.edu
NCSA/U of Illinois at Urbana-Champaign 217-244-0647 voice
605 E. Springfield Ave., Champaign IL 61820 217-244-1987 fax