Feedback on the proposal to change U+FFFD generation when decoding ill-formed UTF-8

Richard Wordingham via Unicode unicode at
Thu May 18 13:03:09 CDT 2017

On Thu, 18 May 2017 09:58:43 +0100
Alastair Houghton via Unicode <unicode at> wrote:

> On 18 May 2017, at 07:18, Henri Sivonen via Unicode
> <unicode at> wrote:
> > 
> > the decision complicates U+FFFD generation when validating UTF-8 by
> > state machine.  
> It *really* doesn’t.  Even if you’re hell bent on using a pure state
> machine approach, you need to add maybe two additional error states
> (two-trailing-bytes-to-eat-then-fffd and
> one-trailing-byte-to-eat-then-fffd) on top of the states you already
> have.  The implementation complexity argument is a *total* red
> herring.

For big programs, yes.  However, for a small program it can be
attractive to have a small hand-coded routine so that the source code
can sit in a single file.  It can even allow a basically UTF-8 program
to meet a requirement to be able to match lone surrogates in a regular
expression, as was once required.


More information about the Unicode mailing list