my machine: class: desktop CPU : AMD Athlon(tm) 64 X2 Dual Core Processor 4600+, cpu MHz 1000.000 cores: 2 RAM : 2 GB HDD : 250GB SATA WD Caviar SE 3.5" 7200rpm ATA300 Hard Drive 8MB Cache OS : Fedora 11 Linux version 2.6.30.9-96.fc11.x86_64 on my machine: * Fig16: - deletes take much less than reported in the paper. the time is almost constant 0.0002 ms, whereas in the paper it goes up when increasing the PDT size. I also do not get the occasional peaks reported in the paper. - modifies and inserts are reproducible. * Fig 17: - VDT (int) takes much longer  on my machine for all scenarios (1M-100M tuples), whereas PDT(int) and PDT(str) are reproducible. - VDT(str) tends to go much quicker off charts in my case as well, the worst case is for 100M tuples when it can be barely seen in the graphic. * Fig 18: - I would say that it is mostly reproducible, the performance for VDT is at times poorer on my machine, the PDT times are very close to those reported. * Fig 19 (TPC-H): - The TPCH experiments look OK. I can therefore support the primary reviewer's report. Dan Olteanu