From gusbroo@fnal.gov Thu Apr 29 17:11:13 2004 Date: Fri, 16 Apr 2004 11:24:16 -0400 (EDT) From: Gustaaf Brooijmans To: Serban Protopopescu Cc: d0dfwg@fnal.gov Subject: Re: Questions at the ADM On Fri, 16 Apr 2004, Serban Protopopescu wrote: > But that could still be applied centrally at the time the root trees are > being made rather then in d0reco. It does imply the root tree is made > in separate steps (by that I mean inside the same program) > but there are no structural impediments to doing that. It does > also imply that algorithms applied centrally cannot be a free for all. True, but I think we should be careful about the number of stages at which we apply corrections. In the ideal experiment, where all your correction factors are determined perfectly during data taking, the perfect reconstruction program applies them all and all the analyzer gets are the final 4-vectors with some tuneable efficiency/rejection parameters. We obviously can't do that, but we should try to limit the number of stages at which we run algorithms. BTW, you seem to imply it's a foregone conclusion that root trees are the final format that will be produced by the farms. You should be careful about jumping to conclusions. Gustaaf > > Serban > > Gustaaf Brooijmans wrote: > > >>I see no compelling reason for an algorithm applied after d0correct > >>that is fast compared to unpacking to migrate (except if the sum of all > >>is substantial compared to unpacking). > >> > >> > > > >I do - the fact that it gets applied centrally and nobody makes a mistake > >when running it! We wouldn't have d0correct if all correction could be > >made at reco time... > > > >Gustaaf > > > > > > -- Gustaaf Brooijmans - Columbia University @Columbia: (212) 854 4527, @Nevis: (914) 591 2804 @Fermilab: (630) 840 4269, emergency: (773) 517 1673