Speaker
G B. Barrand
(CNRS / IN2P3 / LAL)
Description
Rio (for ROOT IO) is a rewriting of the file IO system of ROOT.
We shall present our strong motivations of doing this
tedious work. We shall present the main choices done
in the Rio implementation (then by opposition to what we
don't like in ROOT). For example, we shall say why
we believe that an IO package is not a drawing package (no
TClass::Draw) ;
why someone should use pure abstract interfaces in such package
(for example to open cleanly to various dictionaries) ; how we can
have
a more reliable system than ROOT (for example, by simply protect
the various buffer overflows).
We shall cover the today role of Rio within OpenScientist to store
histograms and tuples. We shall present the effort done around Gaudi,
at the beginning of 2003, to read LHCb events with Rio
(then in the "before POOL" system).
We shall present our views about the LCG proposed solution
for storage, that is to say POOL over ROOT, and why the author
believe that this coarse graining assembly is simply poor software
engineering. We shall explain why CERN, due to its fermionic
sociology,
is going to miss an essential target : an appealing open source
object
oriented data base for HEP. We shall explain then how to do it
without
this lab, then passing from Rio to RioGrande...
Primary author
G B. Barrand
(CNRS / IN2P3 / LAL)