Tracking.

There doesn't appear to be anything under:
http://cvs.gnome.org/viewcvs/*checkout*/orca/docs/doc-set/orca.html#URSPEECH
that specifies when you should spell in military mode as opposed to normal
mode.

How should this be done?

From:

http://www.faqs.org/faqs/radio/phonetic-alph/full/

the phonetic alphabet is:

Alpha Bravo Charlie Delta Echo Foxtrot Golf Hotel India
Juliet Kilo Lima Mike November Oscar Papa Quebec Romeo
Sierra Tango Uniform Victor Whiskey Xray Yankee Zulu

Interesting URL!  The thing I see as being necessary is some way to handle other languages that include more than just A-Z.  For example, the Swedish entry looks like it might have 29 entries. Gnopernicus does its work in speech/libsrs/srs-speech.c, but it seems to limit things to A-Z.  I'm not sure this is really what users want/need, though, since I believe the goal of "military spelling" for screen readers is a bit larger than the original goal of real military spelling.  That is, I believe it is like military spelling in that helps disambiguate characters that may be otherwise confusable when spoken via TTS (e.g., "s" and "f", "p" "t" and "d"), but the screen reader case may cover the complete alphabet for the language.

Created attachment 72994 [details] [review]
Patch to implement this feature.

The attached patch does the following:

* Replaces the util.isDoubleClick() routine with util.getClickCount
  which return the count of the number of clicks a user has made to 
  one of the keys on the keyboard.
* Implement phonetic (military) spelling on a triple-click for
  Numpad-2 (flat review current character)
  Numpad-5 (flat review current word)
  Numpad-8 (flat review current line)
  This includes using the UPPERCASE_VOICE for upper-case letters.
* Fixes up a traceback is _reviewCurrentCharacter() if string was None.

Double-click will continue to ordinarily spell the character/word/line.
Four or more clicks will be treated as if the user had just done a single
click.

Changes checked into CVS HEAD. At the moment it will do
phonetic spelling of a-z and A-Z. Any other characters it tries
to phonetically spell will just be spoken as normal. 

This is on a par with gnopernicus so I'm closing as FIXED.

Oh, the other thing I should add is that I've only tested this 
with gedit where it seems to work nicely.

OpenOffice and Evolution might be another story. If you find 
problems with them, please file fresh bugs and I'll take a look 
at them.

After working with this feature for a while having this functionality available on most flat review keys really isn't useful.  I'd like to have it restricted to the "say-line, say word, and say character keys"  ie numpad 8, 5, and 2.

Just to clarify, where *don't* you want military spelling
that currently has it? 

I ask because the calls to where phonetic spelling is done, 
are in _reviewCurrentLine(), _reviewCurrentItem() and 
_reviewCurrentCharacter() in default.py; methods that are called
from Numpad-[7,8,9], Numpad-[4,5,6] and Numpad-[1,2,3].

It's probably possible to make it work for just Numpad-[8,5,2]
but the code might be ugly.
 

The problem is when you do for example prior or next character really quickly you get military spelling when you really just wanted the next character.  

Created attachment 74842 [details] [review]
Patch to fix the new problem.

The only two places where this was broken was in review previous or
next characters. Review preview or next word or line was doing the 
right thing.

New changes applied to CVS HEAD. Closing as fixed (again ;-) ).