Re: Question on DSIG and head flag 11

There are a number of sub-processes other than the glyf transformation that
do not guarantee bit-identical lossless compression. Table order was
mentioned by the original poster, but there's also padding, and possibly
other subtle details. Supporting DSIG on CFF-only fonts would be
technically possible but would be a fairly major change to the format, and
doesn't really further the use case (web fonts, where signature checking is
not deployed).

I think the original poster might want find that just using a high quality
lossless compression algorithm on the OpenType font meets the requirements.

Raph

On Mon, Jun 29, 2015 at 7:53 AM, Ken Lunde <lunde@adobe.com> wrote:

> Cosimo Lupo.
>
> I was thinking along the lines of processes (encoding/decoding) that don't
> treat the entire font resource as a single entity. I defer to the expertise
> of you and others with regard to this particular concern.
>
> -- Ken
>
> > On Jun 29, 2015, at 7:18 AM, Cosimo Lupo <cosimo.lupo@daltonmaag.com>
> wrote:
> >
> > Ken,
> > If the transform is indeed loss-less, then the output is bitwise
> identical to the original and the DSIG will still verify.
> >
> >
> >
> >
> > On Mon, Jun 29, 2015 at 2:00 PM, Ken Lunde <lunde@adobe.com> wrote:
> >
> > Cosimo Lupo,
> >
> > Pardon the possible naïve question, but wouldn't any transformation of
> the font resource, including non-lossy ones, render the digital signature
> in the 'DSIG' table invalid? I sense a security concern/issue here.
> >
> > Regards...
> >
> > -- Ken
> >
> > > On Jun 29, 2015, at 5:30 AM, Cosimo Lupo <cosimo.lupo@daltonmaag.com>
> wrote:
> > >
> > > Hello,
> > >
> > > The issue of WOFF2 encoder dropping DSIG table for CFF as well as TTF
> fonts the was discussed again in fontTools forum:
> > >
> > > https://github.com/behdad/fonttools/issues/306#issuecomment-116605139
> > >
> > > Chris asked me to re-raise the issue here as well.
> > >
> > > You may recall, back in March, I proposed to keep the DSIG table at
> least for CFF fonts, since these don’t undergo lossy transforms as TTF do.
> > >
> > > In his reply, Vladimir was concerned about the possibility that table
> reordering might occur on either the encoding side (as it’s still the case
> with reference and OTS implementation), or on the decoding side (to comply
> with the OFF recommendations).
> > >
> > > As for the encoder, implementations could be easily modified to allow
> keeping the original table order. Similarly, on the decoding end, the
> latest WOFF2 spec use “must sort” only with reference to the sfnt table
> directory, not the table data order.
> > >
> > > From Adam’s comments in fontTools forum, I gather the reasons for
> dropping the DSIG in WOFF2 are not simply technical ones.
> > > I personally don’t have a strong opinion on the matter, so I’ll leave
> that to you.
> > >
> > > Cheers,
> > >
> > > —
> > > Cosimo Lupo
> >
> >
> >
>
>