Re: [Freetengwar] Telcontar Feature IDs

[<j_m...@sh...>]

On 2016-12-07, at 06:55, mach wrote:

> On 2016-12-06, at 08:47, Arno Gourdol wrote:
> 
> > OK, I have found some contacts at Microsoft and at the W3C. 
> 
> Great news. It's good to have someone around with contacts.
> 
> > Could you send me a test case showing the difference in
> > behavior between MS Edge and Chrome/Safari/Firefox? I'll try
> > to make a case that Edge should be consistent with the other
> > browsers. 
> 
> I will be able to write a test case on Friday or this weekend.

It took me longer than expected. Here it is:

http://freetengwar.sourceforge.net/testcases/font-feature-settings_testcase/

Things are more convoluted than I thought. The good news is
that the behaviour of the Microsoft browsers has nothing to
do with the PUA and that Internet Explorer is no dumber than
Microsoft Edge.

The bad news is that the Microsoft browsers turn off all the
OTL features that are supposed to be turned on by default
according to Microsoft’s OpenType Layout tag registry –
including the GPOS features (which are not included in the
above testcase).

And it gets worse: Some features will be turned on as soon
as there is any "font-feature-settings" entry in the CSS.
Take for instance "ccmp". This is supposed to be turned on
by default. In the Microsoft browsers it is turned off. If
you put 'font-feature-settings: "ccmp" 0;' in the CSS,
"ccmp" is turned on – even though this particular setting is
supposed to do the exact opposite: it is supposed to
explicitly turn off "ccmp". If you put
'font-feature-settings: "dumb" 1;' in the CSS, "ccmp" is
turned on again – even though this particular setting is
pure nonsense and the result should fall back to the
default. However, this nonsense setting will have an effect
in the Microsoft browsers and the result will look different
from the default. Other OTL features that are affected by
this strange bug include the GPOS features "kern" and "mark"
(but not "mkmk") – this list is not exhaustive.

Consequently, the Microsoft browsers can be forced to
emulate most permutations of feature settings, but not all
of them. Something like 'font-feature-settings: "liga" 1,
"ccmp" 0;' cannot possibly be displayed by either Microsoft
browser.

So there are several cases you can make:

* The Microsoft browsers should be consistent with other
  browsers.
* The Microsoft browsers should be consistent with
  Microsoft’s OpenType Layout tag registry.
* The Microsoft browsers should have a way of displaying any
  permutations of OTL features such as
  'font-feature-settings: "liga" 1, "ccmp" 0;'.
* Nonsense entries like 'font-feature-settings: "dumb" 1;'
  should not have any effect in the Microsoft browsers.

As for the other browsers, Firefox and Google Chrome do
exactly what they are supposed to do: They have all the
right default settings and honour all explicit settings,
while nonsense settings do not have any effect.

Safari has some bugs. A number of OTL features is always on
and cannot be turned off, including "ccmp", "kern", and
"mark". Other features appear to be affected by the same
strange bug like the Microsoft browsers, including "mkmk" –
it is turned off by default, but it will be turned on by
'font-feature-settings: "mkmk" 0;' or by
'font-feature-settings: "dumb" 1;'

-- 
grüess
mach