Meeting minutes 2012 02 14

Attendance: Robert, Jason, Jarno

Item 1: 1.6 Milestone 1

Released last week. Need to update main DITA-OT page so that the "Download now" link still pulls 1.5.4, not the latest test build.

Removed a lot of outdated / deprecated PDF code. Added a message for additional deprecated items that will be removed in Milestone 2.

Item 2: 1.6 Milestone 2 plans

3 week iteration ends March 2; code freeze Feb 24.

Plans for M2:

• Jarno: general Java development. Working on a client project that uses PDF2; table processing is hideously complex. Default creates a border around the table, and a line below the header but nowhere else. Only if the header appears on the first page. Strange - doesn't fit into the usual defaults for OASIS tables. The values are hard coded now in many places, so would be good to use variables that can be set easily.
• Also: old bug of when table breaks in multiple pages, bottom border isn't drawn properly. Thinking of fixing that.
• Also: refactoring the way title numbering is generated. Just chapters / appendices. Code uses "get topic type" approach. But it's a single string - need to know if inside <notice></notice> in <frontmatter></frontmatter>, for example. Have an update to simplify, also allows numbering at any level.
• Jason: Continue following bug tracker.
• Robert: Minor tracker items, doc updates.

Item 3: Description of build process

Jason sent a description of the current process to Robert; Robert to post on this wiki. Jason uses RTC to build; found a "default codepage" setting that we hope will eliminate the codepage problems encountered with the 1.5.4 final build. If there are plans to add any new languages to PDF, this would be a good way to test.

Item 4: Other items

Jarno noticed dita-users question about bug in 1.5.4. Should we think about a bug fix release? Doing testing on that will take a week away from 1.6 tester resources. Would be nice if we had some sort of Ant / JUnit based process that could run the tests, do the comparison, and report differences for manual testing.

Current pre-processing architecture is module based. Some Java, some Java+XSLT, some XSLT. The only-XSLT modules use Ant's XSLT task. Could replace with Java wrapper of our own; enables (in server environment) multi-threaded processing. Sounds good.

Discussion of memory issues in move-meta-entries. The XSLT code could be improved by using Saxon functions that do better garbage collection. May need to do the same with the Java code; using the same DOM tree for every topic will waste memory.

Other item: Jarno was playing and converted all of the Ant code to Python. It worked. If we did this, we could get an implementation that runs under .NET, etc. Longer term vision (maybe) of replacing Ant with Python in DITA-OT 2.0 or 3.0, but for now could just be alternate build. Related to idea of making plug-ins more compatible with OSGi so that you could build a plug-in with Eclipse.

Idea for documentation: "using our own dogfood", we should be adding index terms. Also might be nice to add some glossary topics. Robert will do index terms during any doc updates this iteration.