These are good questions, since it is likely that some applications would want to allow input of dashes to be valid, for example.

I think we should focus on the validation of the barcode number itself, since they are the only characters relevant to the checksum algorithm that the validator performs.

Let's assume a simple situation where an application accepts some input string to be used as an IBSN barcode for searching a book catalog. The first thing to do is to filter the string of all but digit characters, because only the digit characters matter for the checksum algorithm. The digits are verifed using a validator before tasking the database with a more expensive search operation on the input. This can be achieved by running Zend_Filter_Digits on the input before validating it (e.g., Zend_Filter_Input supports this nicely). If the database stores the ISBN as all digits, then we have no problems. But let's now assume that the database stores the ISBN as a string with dashes in appropriate places (e.g., "99921-58-10-7"). Another [escaping] filter is needed to transform the data from Zend_Filter_Digits (i.e., "9992158107" needs to become "99921-58-10-7"). A custom filter can be written for this purpose rather easily.

In short:

• dashes are not allowed for input ISBN; they should have been stripped by the time the string is input to the validator
• prefix of "IBSN" is not allowed; any prefixes and suffixes should have been filtered out by the time that the input string reaches the validator