For 8-bit symbols, Reed-Solomon codes are of the form $(n,k) = (255-z,255-2t-z)$ where:
- $n$ is the number of symbols in a codeword
- $k$ is the number of data symbols that can be encoded
- $t$ is the number of errors that can be corrected
- $z$ is the number of implicit zeros that can be used if the code is to be shortened.
What are the restrictions on $t$? In other words, can I choose any $t$ from $t=1$ (for example (255,253)) to $t=127$ (which is (255,1) or 1 data symbol encoded as a codeword of 255 symbosl)? Or is there a maximum $t$ value?
Increasing $z$ increases the number of information symbols to be deleted. So, the maximum value of $z$ is $n-2t-1$ at which point you are left with a $[2t+1, 1]$ "repetition" code that can correct $t$ errors. The word repetition is in quotation marks to remind you that there are $q$ codewords but they do not to need to be repetitions of the same symbol (except for the all-zeroes codeword, of course).
If you allow $t$ to vary, then the original code, before any deletion of $z$ information symbols, is a $[255,255-2t]$ code and so you can have anything from a $[255,253]$ single-error-correcting code (for which $z$ can be chosen to be any number from $0$ to $252$ to get a $[255-z, 253-z]$ code) to a $[255,1]$ $127$-error-correcting code for which $z$ must perforce be $0$.