Compiling the lists has entailed two sorts of work: reconciling existing information contained in IS, J, and P, and striving for accuracy in describing cancels from my own collection.
Reconciling existing information
Using J or P alone to 'type' a cancel is not always easy. J uses a system of general types, to which an individual cancel may not quite conform, and sometimes the distinguishing difference between two of his general types is difficult to ascertain from the type-illustrations he gives. He mentions this in his introduction, using types R4 and R14 as examples, and there are quite a few others. Comparing his data with P's is made more awkward if one can't be sure which J cancel one is to compare with a given one from P.
Because J (Ed 2) took a lot of dates from P, I have presumed that where date ranges coincide, the P and J cancel types also match. Where this does not help I have used any examples I have, together with characteristics such as cancel size, to guess the correspondence.
P illustrates every cancel, which generally makes identifications easier. In a few cases his illustration isn't quite right (for example, it may show an * where there isn't one), which may be because some of his illustrations come from impression books rather cancels 'in the field'. Additionally there is the not uncommon difficulty of distinguishing between cancels which he gives as different. Sometimes a subtle distinction becomes clear on examination, in other cases, even using Photoshop to superimpose cancels fails to detect a difference. It may be that some types shown as separate in impression books are visually indistinguishable in practice, and I have taken the pragmatic view that if one can't tell the difference 'in the field' it has little practical use.
In a few cases it seems likely that P has allocated more than one type to the same canceller which became significantly worn over time. Hopefully the explanation column in the lists makes this clear in the relevant cases. I am certain to have made quite a few mistakes.
P illustrates numerous variants (as D1 State 1, D1 State 2 etc - there are 10 states for Offa D1!), occasionally mirrored in J, but more usually ignored. I have retained all these, believing that, although of questionable historical importance, they add considerably to the fun of collecting.
Where the given usage date ranges don't agree I have generally chosen the wider range unless something strikes me as odd (for example a probable transcription error). The origin of the dates I have given is indicated via the key at the top of each list. Where there is a big discrepancy between dates given I have noted it in the explanation column.
Striving for accuracy
In an ideal world one would be dealing with entire cancels, cleanly struck, as per P's illustrations. In practice the cancel may be partial, faint, smudged, or missing (for example) the year part of its date slug. Technology often helps: the SG Zoom digital microscope has been a real asset, and using Word's power to search lists when faced with a cancel showing only part of a name: for example '..ssa Ig..' turns out to be the middle part of Agbassa Igbadu. Frank Walton's reverse names list on the WASC site can come in handy too.
Accuracy is particularly important when claiming a new earliest or latest use date, a new type or even a new office, and in those cases I have erred on the conservative side. For new dates I have at times relied upon the date of issue of the stamp itself to fix a latest use date, and have said so in the explanation column. Confidently assigning a new type to one of the general types in J can be impossible without having an entire clear cancel (for example, you can't assign a 30mm cancel to type unless you can see the bottom part of the cancel). Again, I have used the explanation column to comment on this.
I have followed the implicit practice of P when deciding whether a cancel is a new state of an existing type: he seems to have assigned a new state for any easily discernable difference. In a few cases there appears to be a continuum of examples between two given types.
In adding a new post office / agency to the list I have often been helped by P who has mentioned its existence without showing a type example. In others, a two part name refers to an identifiable village and its nearest administrative centre. Changes of local governement boundaries mean that the same post office may have been renamed several times, and in these lists each name is given separately. The same applies to the common practice of changing the spelling of a name.
Internet searches and online maps have proved invaluable in tracking down the locations of new offices given in the lists.