W3C home > Mailing lists > Public > public-html@w3.org > April 2008

Re: The mapping of phi

From: Neil Soiffer <Neils@dessci.com>
Date: Wed, 2 Apr 2008 09:22:07 -0700
Message-ID: <d98bce170804020922o1bae44cx5f8d96973b39f1bf@mail.gmail.com>
To: "David Carlisle" <davidc@nag.co.uk>
Cc: elharo@metalab.unc.edu, public-html@w3.org, www-math@w3.org
Banging the drum yet again, there is no way to deal with &phi and &phiv
mappings that "works" for everyone.  No matter what you do, over time as
font designers update their fonts or come out with new ones, the appearance
of these characters will change from a few years ago because font designers
will look to the Unicode spec, not what HTML defines, as to what to draw for
each code point.

As a random guess based on what we found when we looked at fonts that define
these characters, at this point in time, about 2/3 of the pages that use
these characters will look different in HTML4 than they did four years ago.
Updating the HTML5 mappings to the Unicode spec will restore the appearance
that they had four years ago at the expense of the other 1/3 of the pages.
Over time, the percentage of old fonts being used will move towards zero.
If the mapping change were made to HTML5,  that would mean almost all pages
would look like they used to.... except for the ones where people have
changed the entity name because they realized that the name was mapped wrong
and did the obvious work around.

Neil Soiffer
Senior Scientist
Design Science, Inc.
www.dessci.com
~ Makers of Equation Editor, MathType, MathPlayer and MathFlow ~



On Wed, Apr 2, 2008 at 8:55 AM, David Carlisle <davidc@nag.co.uk> wrote:

>
>
>
> > Unicode cannot change a glyph becuase it never assigned a glyph in the
> > first place.
>
> But it can (and did) change the charts at unicode.org swapping the
> sample characters for the straight and curly phi symbols. So when font
> designers are putting the unicode tables into their fonts they can (and
> did, and do) assign the straight phi to one or the other of two slots
> depending on the age of the unicode documentation they are looking at.
>
> This changed _after_ HTML 4 came out.
>
> So, what should html5 (and xhtml, and mathml,...) do
>
> If it keeps phi with the html4 uniocde number then that gives a feeling
> of stability, but if the user is using newer fonts, the visual
> appearance will change, even looking at old pages.
>
> If on the other hand html5 updates the mapping of phi, then for people
> using newer fonts they'll get the intended character, unless they
> actually intended to get the other character (because they looked at what
> their browser gives now, rather than any updated uniocde charts)
>
> Of course there are similar problems for pages that use nummeric
> references or character data, but there is less indirection there so you
> can't do anything about it.
>
> Currently (as far as I know) the mapping is the only difference between
> xhtml1 (and html5) mapping and defined here:
>
> http://www.w3.org/TR/xml-entity-names/
>
>
> The phi problem is flagged in that spec
> http://www.w3.org/TR/xml-entity-names/#diffs
>
>    XHTML uses U+03C6 (decimal 966) GREEK SMALL LETTER PHI, in these
>    files phi is defined as U+03D5 (decimal 981) GREEK PHI SYMBOL
>
> MathML3 will use the mappings derined there by reference.
>
> I think it's essential that HTML5 agrees with the definitio there so
> that it's easy for HTML5 to agree with any XML formats that choose to
> reference that set. So long as HTML5 and XHTML2 groups can agree
> on the mapping of phi, it's easy to ensure that  the xml-entity-names
> agrees with both of them (I'll just change it to match)
>
> There's no right answer, but I think html xhtml, mathml, docbook, etc
> all ought to use the same answer, especially as there keep being
> proposals (for html or even future  xml  versions) just to build in these
> definitions and not need them to be explictly declared, that's a lot
> easier if everyone agrees what the definition should be...
>
> David
>
>
> ________________________________________________________________________
> The Numerical Algorithms Group Ltd is a company registered in England
> and Wales with company number 1249803. The registered office is:
> Wilkinson House, Jordan Hill Road, Oxford OX2 8DR, United Kingdom.
>
> This e-mail has been scanned for all viruses by Star. The service is
> powered by MessageLabs.
> ________________________________________________________________________
>
>
Received on Wednesday, 2 April 2008 16:23:56 UTC

This archive was generated by hypermail 2.3.1 : Monday, 29 September 2014 09:38:54 UTC