The pointless program (see Scaling and Assessment of Data Quality, Philip Evans, Acta Cryst D 62, 72-82) is part of CCP4 and serves to identify space group possibilities from unmerged data.
Typical usage for space group determination is
echo SETTING SYMMETRY-BASED | pointless XDS_ASCII.HKL
One may send several commands to pointless:
echo SETTING SYMMETRY-BASED\\n CHIRALITY CHIRAL | pointless XDS_ASCII.HKL
Typical usage for converting to multi-record MTZ file (for scaling with SCALA or AIMLESS) is
pointless -copy XDS_ASCII.HKL hklout XDS_ASCII.mtz
(the -copy option is not required)
More complicated operations may be scripted:
#!/bin/csh -f pointless <<eof SETTING SYMMETRY-BASED xdsin XDS_ASCII.HKL centre eof
will make an R-factor search for the correct origin.
Why "SETTING SYMMETRY-BASED"? The space group number that pointless writes into the logfile (e.g. "17") is the generic one (as in the International Tables vol A), not the specific one (which could be 17, 1017 or 2017) that is used as a special convention within CCP4 to encode which of a,b and c is the two-fold screw axis. At the same time, pointless re-indexes orthorhombic such that a<b<c : this is what is called "cell-based setting" in the pointless documentation, and unfortunately it's the pointless default. However, users of XDS, XSCALE and XDSCONV want the "symmetry-based setting" because these programs do not understand 1017 and 2017 (and those numbers are not even in the logfile, only in the MTZ file). This also affects space group 18 ("P 21 21 2"), and in space group 5 the distinction between space groups C2 and I2.
This is discussed under the headings "A note on setting conventions" and "SETTING CELL_BASED || SYMMETRY-BASED || C2" in the pointless documentation.
What can go wrong when using the default? For example, the space group assignment of data sets differing a tiny amount in two cell parameters may flip between two alternatives, i.e. the assignment is not stable. E.g. two datasets from a P2221 crystal with axes of 100,120.5,120.7 and 100,120.7,120.5 would be considered (and re-indexed!) by cell-based assignment as P2221 and P2212, respectively, with ugly consequences for data merging and all downstream steps of data usage. The example shows that enforcing a<b<c would be considered more important than assigning the spacegroup in a reliable way, which makes little sense.
The upshot is: for space groups 5, 17 and 18, XDS users want to use the "symmetry-based setting", and it does not hurt to always use it, i.e. for all 65 space groups that proteins crystallize in. This deviation from the default of pointless has several advantages (avoiding the ambiguity, and avoiding change of space group assignment upon small change of axes), and no disadvantage.