ICU meeting 2006 Apr 19 Copyright (c) 2006 IBM and others - Attendees: - Mark, Markus, Vladimir, Deborah, Doug, George, Ram, Andy, Eric, Michael Ow (IBM), Eric, Steven - Copyright Assignment - discussed, sent to list. All to discuss with lawyers. - RBBI French - #1574 - There are probably some other bugs related to this, italian (#3917), etc - Suggestion: Transfer bug to CLDR, close in ICU - Deborah has the action to open a CLDR bug and close the ICU bug. - UText - iteration boundaries - Issue: May want to iterate over codepoints between start and limit. Would be convenient if the iteration code didn't have to call getNativeIndex at each point. - Question: implement this without the provider being aware of it? - Proposal: add field for start/end offset and keep it reserved - modify [to point at] chunk boundaries in access function - Proposal: have two areas of reserved data in UText: One the providers won't touch (UText's use) , and one that providers can use as their scratch area, initialized to zero. - Andy to rough something out. - This current feature might want 20 bytes, make it 32 bytes? - Or, better to heap allocate any extra stuff. - Final proposal: Pointer (for the provider?) , and some extra area for UText - RBBI/UText question - Andy - there is an API for getting the old CharacterIterator. Doesn't work so well with UText.. - Can we wrap a CharacterIterator in a UText? yes - UText wrapped in CharacterIterator isn't fully implemented - would be a pain. - Concern about burden of learning on UText providers -- Mark - There are two common cases, which could help providers - #1 UText user knows its going to use the entire buffer - (provider could make one big chunk instead of pieces) - #2 UText user knows it's not going to use native indexes often - (provider could not provide the index map) - Provider COULD statistically analysize... - It's better, for the user, to not have to make choices.. - Proposal: hint bits from the user to the provider, to request certain types of behavior. - Can the native index map be lazilly created? - Andy would like to try different things with UTF-8, and see if hint bits are justified. - Summer of Code - IBM: Q: RBManager? A: not interesting for Google -- May need to make sure it is an item which will integrate back into ICU. - Tradeoff: handholding has to be done. So, someone from the team has to be a mentor - sponsor. Need to ask in each manager chain for availability to mentorship. - Need someone to put together a formal list - Students to start applying May 1st. - Apple: Lenient date parsing. Might be beyond what would work. - Google: State table based Dictionary lookup? Mark to write something up. Deborah's CTD is optimized for match vs. no match. -