ICU development team meeting Call-In: 1-888-790-7707, 1-415-228-4574 Date: 2000-May-09, Tuesday Time: 9am PDT - 10am PDT Place: IBM Java Technology Center 10275 North DeAnza Blvd. Cupertino, CA 95014, USA. Host: IBM ======================================= Attendees: Jim Snyder-Grant (Lotus) Mark Davis (IBM) Markus Scherer (IBM) Tex Texin (Progress) Steve Watt (Progress) Steven Loomis (IBM) Vladimir Weinstein (IBM) Minutes Taker: Markus Scherer Agenda: - Roll call See attendee list above. - Additional agenda items - Action items from previous meetings See action item list below; comments from today are marked with today's date. - ICU 1.4.1.2 is ready to be released! This is an update to 1.4.0 and 1.4.1.1 for DB/2 which is starting to use the XML parser with ICU underneath. Some of the issues were correct default encoding names and default locales, as well as finding and loading ICU data, which prompted a modification of udata and, in the current code for 1.5, a rewrite. - ICU 1.5 release milestones: (GA 6/7/2000) Currently scheduled items to be finished for the 1.5 development release: - Finish Bidi API implementation - Clean up callback API for converter - Add 3/4 byte encoding support,e.g. euc-jp - Complete charmap generation program on various platforms - Define locale data XML format spec - Provide integrated tools for data generation - Support IANA/MIME names (Yves?) - Complete RB/Collation data integration - Integrate MP-RAS port changes. - Complete upgrading to Unicode 3.0, Normailzation. - Provide basic collation versioning. - Complete transliterator rule syntax update. - Add basic UTF-16 support. - Timezone and formatter clean-up, especially Year-Week-Day handling. - Finish ICU User's Guide Release schedule: 5/22 Implementation Complete 5/26 Code Freeze (including unit tests) 6/02 Test Exit (Items 1, 2, 3, 5, 8 and 9 as well as UTF-16 compatibility, etc.) 6/7 Final Release Date - Time for the agenda items that were added in the meeting + Jim sent an email this morning about opening a converter with a locale as parameter. Discussion: * Lotus needs to specify the locale at open time for the algorithmic LMBCS converter; this is also useful for ISO-2022-JP etc. * Steven pointed out that this would be useful for the other variants of ucnv_open() but that that means to add several such functions. * There should be a clone() function for converters so that one does not need to know the UConverter structure and memcpy it. * The original idea for the locale parameter in the internal open() functions was to use it for ISO-2022. The parameter was to be extracted from the canonical name in convrtrs.txt. * We discussed having a number of parameters as part of the converter name, which could then also be in the canonical name in convrtrs.txt. We need to design a syntax for this. Based on this, we do not need new ucnv_open() variants. + Codepage mapping table mailing list * Mark has started an email list hosted by unicode.org for discussing codepage mapping tables and related issues. * Several ICU members are signed up already or asked Mark in the meeting to sign them up. * We (Helena/Mark) have written a program for Win32 that uses the platform conversion APIs to write mapping tables that precisely reflect the platform mappings. *** Action for all! *** * We need others to port this to other platforms and to generate more such mapping tables. + Collation versioning * Agreement that the version number must be increased at least when the collation keys would change compared to the previous version. * A simple implementation for the upcoming release could just always return "1" or similar, until the collation actually does change. * We should generate a hash code during the compilation of the rules into a binary collation structure. This hash code needs to cover all rules. + Reimplementation of udata for 1.5 * Due to continued problems in the DB/2 environment that had originally triggered the udata mechanism, the implementation was redone. Now there is no compile-time configuration between DLL loading and memory-mapping any more. Instead, both variants are always active. The build options only control the type of data file that is built. * Common data files are first attempted to be loaded from a DLL, then mapped as a single file. Extension is done only with single-data memory-mapped files, not single-data DLLs. DLLs may contain either a single entry point with the same structure as the flat, mappable common file, or per-data entry points as before. In the latter case, we are now also building a Table of Contents that differs from the one in the flat file: it needs to use pointers instead of offsets. * udata_setCommonData() is available and detects either type of ToC. It also auto-detects if the data is prepended by a double dummy as generated in the .c files from genccode for alignment control. The 1.4.1.2 implementation relies on the mapped-file ToC. *** Action for all! *** * Building the ICU data DLL is much faster if it is done via the flat common file by directly transforming that flat file into a .o (.obj) file and linking it. This is done by gencmn now for Win32. Some platforms use the ELF format and have a libelf.sl or similar that allows to write such files. Anyone who is interested in fast builds of data DLLs on non-Win32 platforms needs to write code for the transformation of static const, read-only data into a .o file and add that code to gencmn. - Future meeting schedule Next meeting: 2000-may-23 9:00 PDT - Action items from this meeting See action item list below. Action item status: [r] : removed [-] : completed [ ] : open Mark suggested to order the items by due date. This is done here. Items that are past due are still in the list with their old due date. --- Items that are due by the next meeting --- [-] T20000208-03 [Brendan] Send out the requirements with regards to Collation versioning to icu mailing list and IBM will revise the current versioning proposal to satisfy the combined requirement. Due date: 2000-apr-25 2000-apr-25: Brendan sent out email regarding the discussion to be taken place this week. 2000-may-09: general agreement, and more discussion today about a simple implementation in time for 1.6 [ ] T20000229-03 [IBM & Yves] Improve readme and related documents + Cross-link javadoc texts, C <-> C++ etc. + update readme & docs/* + move coding guidelines to separate doc (only link to oss server page? remove irrelevant items) Due date: 2000-apr-30 2000-apr-11: include a list of files that need to be installed [-] T20000425-01 [Steven L.] Send out more description on the udata_setCommonData() proposal. Due date: April 30, 2000 2000-may-09: Implemented in 1.4.1.2 and in current code for 1.5. [ ] T20000321-01 [David] Submit ICU port changes for NCR NP-RAS platform. Due date: 2000-may-09 2000-apr-25: pending on Steve Watt's feedback on various compiler options. 2000-may-09: Tex/Steve: changes for a very similar platform are checked in and should cover much of this platform [ ] T20000321-04 [Roger] Complete the Conversion chapter of the ICU user's guide. Due date: 2000-may-09 2000-may-09: The chapter iss not ready yet. [ ] T20000321-06 [Yves] Review the Conversion chapter of ICU user's guide. Create draft of the build/installation instructions of the Introduction chapter of the user's guide. Due date: 2000-may-09 2000-may-09: The chapter is not ready yet. [ ] T20000411-02 [Steve W.] Collect the makefile flags for different platforms Due date: 2000-may-09 2000-may-09: Done for UnixWare. [ ] T20000229-05 [David] Review the Format User's guide chapters Due date: 2000-may-09 [ ] T20000229-06 [Tex & Steve] Review the Collation and ResourceBundle User's guide chapters Due date: 2000-may-09 2000-apr-11: there are still broken links in the user's guide on the web 2000-may-09: The links are fixed. [ ] T20000321-05 [Brendan] Review the Conversion and Transliterator chapters of ICU user's guide. Due date: 2000-may-15 2000-apr-25: Feedback was submitted through email, 2nd review cycle is required after the update. 2000-may-09: in 2nd review cycle now [ ] T20000229-04 [David] David Craig will continue to work on the DateFormat performance and report in the next meeting. Due date: 2000-may-15 --- Items that are due after the next meeting --- [ ] T20000411-03 [Markus/Steven/Yves] Document how to drop/add/delete resources (data files) like converters, locales, and collation tables Due date: 2000-june-9 [ ] T20000125-01 [Jesper] Implement the global new/delete operator to use uprv_malloc and uprv_free functions. Due date: August 31, 2000 [ ] T20000411-01 [IBM] Investigate stylebook output in UTF-8 Due date: August 31, 2000 [ ] T20000321-03 [IBM] Find information on the freeware that can print the html pages in a directory. Due date: September 1, 2000 2000-apr-11: we need to investigate how to produce one single HTML file or a PDF document