The specification says that embedded subset fonts must have their names prefixed with six uppercase letters plus "+". For example ACBCDEF+Arial. In every PDF document I look at, the six characters are fairly random sequences. Is there any reason not to just do something much simpler, such as "AAAAAA+Arial", "AAAAAB+TimesNewRoman", etc.? IOW, do the semi-random prefixes I see elsewhere really add any value?
The only thing I can think of is that maybe the prefixes need to be unique not just within a single PDF, but across multiple PDFs. The specification hints at that when it mentions merging PDF documents with incompatible font subsets, but I don't know whether that is actually an issue? So far as I can see the specification doesn't actually require uniqueness of prefixes at all, except to distinguish between two subsets of the same font within a single PDF.
My guess is that the prefixes I see are derived from the set of characters in the subset (some kind of checksum/hash), which would give a reasonable chance of two different subsets having different prefixes. But if that is the case, surely there should be a standard (documented) way of deriving the prefixes, to ensure that different PDF generators don't inadvertently use the same prefix for different subsets?
I just had another thought - are the prefixes maybe used by printers that cache embedded fonts to know that they can re-use a cached font subset? If so, that would effectively require that the prefixes are unique across multiple documents.
The only thing I can think of is that maybe the prefixes need to be unique not just within a single PDF, but across multiple PDFs. The specification hints at that when it mentions merging PDF documents with incompatible font subsets, but I don't know whether that is actually an issue? So far as I can see the specification doesn't actually require uniqueness of prefixes at all, except to distinguish between two subsets of the same font within a single PDF.
My guess is that the prefixes I see are derived from the set of characters in the subset (some kind of checksum/hash), which would give a reasonable chance of two different subsets having different prefixes. But if that is the case, surely there should be a standard (documented) way of deriving the prefixes, to ensure that different PDF generators don't inadvertently use the same prefix for different subsets?
I just had another thought - are the prefixes maybe used by printers that cache embedded fonts to know that they can re-use a cached font subset? If so, that would effectively require that the prefixes are unique across multiple documents.