Ftp-WG: FEAT UTF-8 in review

> Clients & servers should also liberally accept LF as
> Telnet end-of-line, yes?

The only eol talked about for telnet are CRLF and CRNUL, the telnet allows
others as local options.  The problem is you don't know what is eol, except
that it's not going to be one of the 96 graphical ASCII characters.

> > When transmitting UTF-8 encoded characters,
> > the shortest form should be used.
>
> http://www.unicode.org says Must use shortest form, to
> avoid the hole of a s*curity filter rejecting only
> shortest-form and failing to reject a longer, not so
> plainly equivalent, form, such as the UTF-8 outside of
> *( %x20-7E / %x80-EF ) that isn't shortest form for
> any char of the x0000..FFFF range.

From the point if view of the FTP, 'should' is good enough.  Let's allow
that there may be good and valuable reasons for some applications to ignore
the requirement.

> > The ABNF for pathnames presented in [RFC2640] is
> incorrect.
> > ... the correct syntax is
> > PATHNAME = *( %x20-7E )
> > PATHNAME = *( %x20-7E / %x80-FF )
> > both cases render moot all discussion
> > about the use of the characters CR, LF, and NUL, in
> pathnames.
>
> Eh?

> As a server, in a folder, I have a local filename that
> includes a local Eol.
>
> How should that filename appear in my NLST data,
> supposing the client asked for FEAT UTF-8 NLST?

It cannot appear on ANY FTP command or response.  Sorry.

> If that conclusion is correct, should we state it?  If
> the shortest form UTF-8 encoding at the client or at
> the server contains %x00-1F, then those bytes should
> be transmitted as the next shortest form i.e. x C0 80
> .. x C0 9F?

I avoided the issue in this draft as being too contentious.  I'm happy if
people just realize that you cannot send ( %x00-1F / %x7F ) on ANY FTP
command or reply and expect correct interoperation.

> > OPTS <SP> UTF-8 [ <SP> NLST ] <CRLF>
>
> Why is NLST an option?  Why not require it in the
> syntax.  Why not say every path name is UTF-8 or not?
> And what about LIST and MLST and MLSD and so on?

I only specified NLST because I don't see LIST as being useful for automata,
and MLSx already seems to handle the issue.