well - i wanted to start by spliting those offsets into the 6 groups for each byte
a bit of luck - that went surprisingly well :t
because of the fact that bytes have 8 bits, we can split the low-order base-210 into a base-6 digit and a base-35 digit
that way, we can use the base-6 digit to address 1 of 6 bytes and the base-35 digit to address bits
we can also use the base-6 digit to select the "parse" values for the base-35 values
you can do that by a set of look-up tables or by branching, i guess
nonetheless, when the offsets are split into 6 groups of 8....
------------0
001 01
011 11
013 13
017 17
019 19
023 23
029 29
031 31
------------35
037 02
041 06
043 08
047 12
053 18
059 24
061 26
067 32
------------70
071 01
073 03
079 09
083 13
089 19
097 27
101 31
103 33
------------105
107 02
109 04
113 08
121 16
127 22
131 26
137 32
139 34
------------140
143 03
149 09
151 11
157 17
163 23
167 27
169 29
173 33
------------175
179 04
181 06
187 12
191 16
193 18
197 22
199 24
209 34
that worked out nicely

so, it would seem that the best approach is to actually mix bases
a simple example...
high order low order
base4294967296digit base35digit
in this case, we need to take the base-4294967296 digit and MOD 6 to figure out how to parse the low-order base-35 digit into bits
rather than constantly dividing to get the mod-6 "remainder", we can keep another high-order digit in parallel
high order low order
base4294967296digit
base35digit
mod6digit
the base-4294967296 digit can be used to address bytes
and, the mod-6 digit can be used to parse the low-order digit
the limits of the above method are
max file: 4gb
range: 140g = 150,323,855,360
if we wanted a larger range, we could use a real mixture of bases - lol
high order low order
base4294967296digit base6digit base35digit
max file: 24gb
range: 840g = 901,943,132,160
which makes it a little tricky to address bytes, but it's do-able