RE: suggested WOFF changes

On Tuesday, June 22, 2010 4:41 PM John Daggett wrote:
> 
> Vlad wrote:
> 
> > > > User Agents MUST NOT permanently install fonts delivered in a
> WOFF
> > > > format as system resident fonts, and SHOULD only use downloaded
> > > > fonts to render the content of a webpage that WOFF resources are
> > > > associated with.
> > >
> > > This is redundant, the CSS3 Fonts specification already defines
> this
> > > behavior for *all* font types, not just WOFF [2].  See section 4.1:
> > >
> > >   "Downloaded fonts are only available to documents that reference
> > >   them, they must not be made available to other applications or
> other
> > >   documents."
> > >
> > > The primary reason for this is security, the content of a given
> page
> > > should not influence content of a different page unless the
> resources
> > > are explicitly shared (i.e. the pages link to the same resource).
> >
> > I found an interesting discussion where WOFF was mentioned [1], and
> it
> > appears that the UA behavior/requirements specified by CSS spec with
> > regard to downloadable fonts may not be supported by some browsers.
> In
> > light of this discussion: taking into account that implementers
> expect
> > to see any relevant requirements clearly mentioned in the spec and
> that
> > the WOFF spec is so far the only web font specification developed by
> W3C
> > - I think it's worth to mention explicitly what the expected UA
> behavior
> > must be when consuming WOFF resource, and appending the proposed text
> to
> > the second paragraph of the Introduction section seems to be logical
> and
> > appropriate.
> >
> > I don’t think it would be a problem reiterating what CSS spec already
> > says (and we can also make a reference to CSS spec here to connect
> the
> > dots).
> 
> I'm assuming the conversation here is the one below.  I don't think
> what
> you say follows at all from that conversation, they're talking about
> where in the software stack to implement a WOFF handler, not about
> browser behavior.  What correct user agent behavior is violated here?
> 

In http://krijnhoetmer.nl/irc-logs/whatwg/20100303#l-194, these lines in particular triggered my attention:


<paul_irish> othermaciej: earlier on the WOFF topic.. i just wanted to clarify.. you said it'd be easier to just make woff fonts installable on mac os, right?
<othermaciej> paul_irish: it is simpler for us to treat them like any other font than to try to make the font system only handle them for WebKit
<paul_irish> roger that. i'm just kinda wtf'ing because i thought the intention of the woff format was to avoid a trivial-to-install scenario
<othermaciej> I'm sure that is some people's intent, I don't think there is any such requirement in the spec though

My understanding (I may be wrong though) is that "make woff fonts installable" and "treat them like any other font" doesn’t seem to be compliant with @font-face rule description that says that "downloaded fonts must not be made available to other applications ..." (where installed fonts would in fact be made available).

Also, the last sentence in that fragment clearly points to the fact that right now WOFF spec is silent about it, and I think it is useful to clarify that the expected compliant UA behavior is to treat fonts downloaded in WOFF format as a temporary resource, and this behavior is in line with CSS spec requirements.

> I think the concern here is that *if* Apple makes WOFF a system-level
> font format that it will be possible to drop WOFF fonts into a font
> folder and use them as normal system fonts.  That doesn't make much
> sense to me, no one is producing fonts for this purpose and Apple
> already has a variety of font packaging mechanisms better suited for
> platform font use.  What Maciej is saying is that it may make more
> sense
> to implement WOFF support in a system-level API.
> 

IMHO it doesn’t make much sense either. Once UA decodes the WOFF file structure and reconstructs the font packaged in it - the system-level API would be available and can easily be used to handle a font, just like any other application would do this, e.g. with embedded fonts.

> Duplicating wording across specs means you always need to keep it in
> sync and that you put it in the correct context.  I think it's a much
> better idea to reference wording in the CSS spec and make changes there
> if you feel there's something missing.

This is in essence what I am proposing. Since the relevant wording of the CSS spec isn't easy to reference directly ("see second paragraph after Example VII presented in section 4.1" would be a rather cumbersome reference) I propose to just add one sentence in the WOFF spec and reference CSS spec as origin.

Best regards,
Vlad

Received on Tuesday, 22 June 2010 21:40:20 UTC