// Copyright (c) 2006 IBM

Meeting Feb 1 2006 10am
-----------------------

Present:
Phone: Deborah, Eric, Ram
Cupertino: Markus, Andy, Vladimir, George, Steven (scribe)

* Meeting change schedule:   
   end biweekly PMC meetings
   change to weekly technical meetings with all PMC members invited.

* Discussion of ICU Charset detection port to C/C++
  C is preferred API
  C++ API is there also, but not reviewed?
  Decision: only C API, don't make C++ public.

  Java API: can give back a Stream
  C API: gives back just an encoding.   ICUIO integration isn't done yet, some modularity issues.  Suggestion to put it into the charset detection API instead of the icuio.
  Eric to file RFE to add stream supportĘ(icuio) to charset detection.

  Issue: Reorganization of data to rsrc bundles. sync with J.  Eric to File New RFE and xref.
  Next week: review

* Hide draft?
  Ram:  four kinds of APIs:
   Macros, #defines, structs/classes, functions
  We have HIDE_DRAFT for functions, but harder for enums (compilation failure).  For enums we can ignore them, or try to find another way to handle this.
  Issue: even if the draft enums aren't used, may get a compilation error.   with functions at least you'd get a link error, but in this case you could get a compilation error.
  Would be better if opaque definitions (such as URegistryKey) wasn't hidden.   
  Also difficult for structs, because if an item goes away, the struct layout changes.
Decision: have explicit values for all enums. Ex:
   URBK_TITLE=4
 Then, it can be #ifdef'ed out.

 For bracketing out a whole enum, could bracket around the entire enum, or, bracket around each data item (except for the Count).

* #5025  *scanf* functions return double for %f, etc.
   Summary:  Change, but post to list to verify.

* #4998  Should maxFractionDigits trump MAX_DBL_DIGITS?
  A1: No, don't allow users to set whatever they set.

  Q2: Should the max allowed be MAX_DBL_DIGITS+2?
  A2: can test on i5/OS

  Issue: this can change how the tests work.

  Already have tests disabled for IEEE754 irregularity. 

  Could have configure check for possible behavior.  Otherwise, have an #ifdef in the headers for platforms with known issues.

* #4889  RFE: split some UText providers with new UTextBytes interface
 Defer.

* #3464  hex escape transliterator needs to understand \xhh
  Decision: Remove discuss, keep small.   Make consistent with unescape().    ( means: \x and \U )
  Filtering takes care of "ignore CR / NL". 
  Use transliterator "variants" for options to ignore "\n\r" etc.
  Consider variants for options like "emit \n, emit \r" etc

* #3898 skip (-mark -doug)

* #4128  Choosing ja calendar changes time formats as well
  CLDR data is wrong, Deborah will file a bug. (japanese, etc have timepatterns and shouldnt)

 Issue: ICU fallback handles datetimepatterns as a block.
  Need to change resoruce format.
      Make a table, Split into two, or null/0 length out strings which fallback

  (null strings for fallback): 
        japanese{
           DateTimePatterns{
               "", // fallback
               "",
               "",
               "",
               "EEEE, MMMM d, y G",
               "MMMM d, y G",
               "MMM d, y G",
               "",
               "",
           }       
  [or]
        japanese{
           DateTimePatterns{
               int:0, // fallback
               int:0, // fallback
               int:0, // fallback
               int:0, // fallback
               "EEEE, MMMM d, y G",
               "MMMM d, y G",
               "MMM d, y G",
               int:0, // fallback
               int:0, // fallback
           }       

 srl and ram to work on it.

* URegistryKey
  both @stable and @internal in the same place.
   brkiter: @stable

  Is used in stable API.
 Conclusion: Make it @stable. Move to @internal.

* Q: Freezable?