> How about this:
>
>         movfw   COMND   ;get you command into w
>                 addwf   PCL             ;add offset to pc to generate
>                                 ;computed goto
>
> Make sure you never give COMND a command value higher than the number of
> goto
> statements you have!

I prefer to use [assuming, e.g., 28 commands]:

        movf    COMND,w         ; "Official" form for movfw, though movwf is
                                ;   nice if your assembler will take it.
        andlw   31              ; Next power of two over number the number of
                                ;   commands, minus one
        addwf   PC              ; Or addwf PCL
        [the 28 commands]
        goto    Oops
        goto    Oops
        goto    Oops
        goto    Oops

If the number of commands is slightly over a largish power of two, then I
spring for the compare-against-maximum, but otherwise the andlw is cheap
insurance (one cycle, and one byte for it and every "goto Oops" that's
needed).

If the number of "goto Oops"s would be objectionable and my program needs to
have any data tables within it, I sometimes do this (assuming I have a 40
entry command table, a 20 byte data table, and an 8 byte data table)

CommandParse:
        movf    command,w
        andlw   63
        addwf   PC
        [the 40 commands]
Data1:
        andlw   31
        addwf   PC
        [the 20 RETLW's for the first table]
Data2:
        andlw   7
        addwf   PC
        [the 8 RETLW's for the second table]
        retlw   0               ; In case Data1 was called with 30 or 31
        retlw   0               ; In case Data1 was called with 30 or 31

Using this approach, an overrun on the commandparse table will result in an
RETLW [something] being executed in fairly short order, as will an overrun
on either of the data tables.  The extra code required, however, is nothing
beyond the "andlw"'s and the two extra RETLW's at the end.  Not at all bad
given the level of bullet-proofing it achieves.