Wouldn't something like that be possible; that would also leave out the implication that there's just one <file/></file> pair

To make the problem clearer:

• we have a text like
  <file fiop="fiop_name" length="333333"/>#333K binary chunk goes here#}</file>
• the "binary chunk" may have any length and may contain any data, possibly (with a lower probability) even the ending tag ... \x02\xc5</file>\x64\xf4  ...
• per string $string, there is more than one of such sequences <file .../> ...</file> to be expected

The only chance I'd see here would be to advance to the start of data, extract the data by substr($string, pos($string), $length) and update the string's pos($string) behind the data: pos($string) += $length. At that point, it could be checked for the expected ending tag </file>. All this happens in a while loop under /g until no more <file> blocks can be found.

Could the above text describe problem and solution?

Regards

mwa