What you (@agenta) should consider is that this "floating technique" in theory is efficient/meaningful/clever only if the (different) charsets you are using (in your specific case there are 2, ?a and ?l) are disjoint (see https://en.wikipedia.org/wiki/Disjoint_sets).
The problem is that ?l is a subset of ?a, i.e. ?a contains every char from ?l. They are overlapping. So for instance it would make sense to "float" ?s within a mask of all ?l (e.g. ?s?l?l?l?l?l?l?l, ?l?s?l?l?l?l?l?l, ?l?l?s?l?l?l?l?l, ?l?l?l?s?l?l?l?l ….. ?l?l?l?l?l?l?l?s), but not to "float" ?a within ?l.
This of course is only theory, there are some specific cases (with fast algorithms) where it is more efficient/faster to use a mask for which the total keyspace must be large enough to reach full acceleration (especially the left-hand side shouldn't be too small in general). This of course depends on which algorithm and attack mode is used.
BTW: in your case it would probably make sense to use ?l, ?u and ?s and something like PACK's policygen ( http://thesprawl.org/projects/pack/#policygen )
PS. it is still not very much clear to me what you mean by "special character", the definition of oclHashcat/hashcat of special character is ?s = !"#$%&'()*+,-./:;<=>?@[\]^_`{|}~ (including the space " " character), there is also ?b which is even larger (?b = 0x00 - 0xff), but again you shouldn't "float" overlapping (character) sets, so you could define your own with --custom-charset1...--custom-charset4, e.g. every "meaningful" character NOT in ?l and ?u.
Above, I assumed that the special characters you were referring to are all contained within ?s.
update: I just found out that there was a small problem with policygen.py and it did not respect the --maxdigit value (see fix here: https://github.com/iphelix/pack/pull/2 ), you should either wait for a new fixed version or use the patch (or my modified github master branch of PACK).
The problem is that ?l is a subset of ?a, i.e. ?a contains every char from ?l. They are overlapping. So for instance it would make sense to "float" ?s within a mask of all ?l (e.g. ?s?l?l?l?l?l?l?l, ?l?s?l?l?l?l?l?l, ?l?l?s?l?l?l?l?l, ?l?l?l?s?l?l?l?l ….. ?l?l?l?l?l?l?l?s), but not to "float" ?a within ?l.
This of course is only theory, there are some specific cases (with fast algorithms) where it is more efficient/faster to use a mask for which the total keyspace must be large enough to reach full acceleration (especially the left-hand side shouldn't be too small in general). This of course depends on which algorithm and attack mode is used.
BTW: in your case it would probably make sense to use ?l, ?u and ?s and something like PACK's policygen ( http://thesprawl.org/projects/pack/#policygen )
Code:
python policygen.py --minlength 7 --maxlength 8 --minupper 1 --maxupper 1 --mindigit 0 --maxdigit 0 --minlower 5 --maxlower 6 --minspecial 1 --maxspecial 1 -o my.hcmaskPS. it is still not very much clear to me what you mean by "special character", the definition of oclHashcat/hashcat of special character is ?s = !"#$%&'()*+,-./:;<=>?@[\]^_`{|}~ (including the space " " character), there is also ?b which is even larger (?b = 0x00 - 0xff), but again you shouldn't "float" overlapping (character) sets, so you could define your own with --custom-charset1...--custom-charset4, e.g. every "meaningful" character NOT in ?l and ?u.
Above, I assumed that the special characters you were referring to are all contained within ?s.
update: I just found out that there was a small problem with policygen.py and it did not respect the --maxdigit value (see fix here: https://github.com/iphelix/pack/pull/2 ), you should either wait for a new fixed version or use the patch (or my modified github master branch of PACK).