approval procedure

[Marcus G. Daniels]

Igor Pechtchanski wrote:

>It might help if you included some info about where a particularly obscure
>package was leading (e.g., I have no idea what HDF5 is or does, but I may
>have heard of the other package you intend to produce that requires it).
>  
>
HDF5 is a library for fast and efficient storage and recovery of 
dimensional data.   Unlike SQL RDBMs, it is mainly concerned with the 
portable and efficient storage of large datasets, often of many 
dimensions (space, time, temperature, etc).  Although the kind of 
queries that can be made are a subset of what one can do with an RDBMS, 
it is much faster, typically keeping the data in the native 
representation of the machine (suitably annotated).  Then if the HDF5 
file is given to someone else, the library automatically converts to the 
right representation via the API.  There are many features in the 
library including non-trivial data structures like ragged arrays and 
compound data types, support for different pluggable compression 
techniques, ability to operate `over a wire' in cluster computing 
environments, etc. but the basic idea is that is a portable way to have 
fast, random access to big chunks of dimensional data.   (Something you 
don't have with a text file, say.)

>It might help if you included some info about where a particularly obscure
>package was leading 

The other package I'm working on is a simulator for agent-based modeling (http://www.swarm.org).  HDF5 is useful to this package because it provides a way to record the evolving state of a simulation and to load different sets of initial conditions (e.g. geographical maps).

Thanks for the tip about the package-server.