Meeting minutes 2013 06 04

Attendance: George, Radu, Eliot, Kris, Robert, Jarno, Jim, Matt, Eric

Item 1: Recent development status status

Not too many commits lately. Jarno focus on 2.0, Robert / Eliot / Kris focus on OASIS items.

Item 2: Preprocess Next

• Jarno trying to get to state where can commit to main repository, so that can start to test / add functionality that people need. Big change is treating files as URIs, also XSLT 2.0 work. Other optimizations, rewrite, mostly gen-list and debug-filter have been modularized. Brief email discussion with user about new client in version 2 develop branch.
• Would like to get ready for checkin this month, not sure how much time is available. Not trying to make it 100% compatible / flawless in first pass, so may be better to try to get in, start experimenting. Code has been available in personal hub for 6 months or so, but more likely to have people look if in main hub. Need to work on some of copy-to, once that's done, may move over.
• Eliot wonders if DITA for Publishers stuff would be useful as part of rewrite; Jarno - not rewriting all, just converting to use the Job Object in a better way. Keeping existing XSLT more or less as-is, just updating to "version 2". Once that's in, can start to migrate to use some of those utilities. For example, path2proj PI, start passing in from outside rather than using PI.
• Jim curious about URI - using built in Java stream mechanism for resolving content? Jarno: can use various methods to retrieve content, could use https, could store in semi or fully parsed form. Basically in sum, trying to get rid of Ant - still used to define order of operations, but leave you more freedom to do else. Jim: We integrated OT internally to our CMS, first thing we did - install custom resolver for XSLT engine, that resolver is call back. Also: we cache the XSLT's, so don't reload. For instance if processing 1000 topics in loop, don't reload. Other thing: used pipeline where tried to have generators / transformers / serializer nodes, they implement sax, allowed us to chain together each step. Then just hacked readers / writers because didn't like that they write their own. Didn't like the writing code, so just used a serializer. Jarno: yes, that's the approach we're taking now. Slowly removing code that uses Java writers, use Sax pipe / transformer to serialize. Problem is, can't create one pipe to push all through and ends with single serializer, because some parts in preprocess - need to process all first, then start again. Jim: yes, we broke down into simple idea that entire processing is called Job, got rid of Ant, each step has multiple jobs, preprocessor has 4 steps, in each step, iterate to loop over topics. Think pretty well emulates OT model. Each job step can have iterator, ends with Sax pipe. With that, could do all in OT, don't change code that much. Jarno: yes, that's basically where want to take it. Current way, using Ant to drive all code, works but not great approach. Old gen-list and debug-filter, now rewritten as single step, files only read once / processing is single Sax pipe, can plug in your code / remove some filters. Can use same approach elsewhere, will be able for example to combine key processing with coderef, could be single pipe, can disable some if a doc doesn't need that filter. Same approach, basically. Jim: sounds like good optimization. Problem for us with internal integration is plugins, not compatible with plugins anymore. Jarno: yes wondering at that solution because plugins mess everything up. In new preprocess, extending so you can hook into a given point, doesn't need to be more difficult, but doesn't have to be as easy as for new transform types. New preprocess, can optimize, don't need exact same hooks in same places because old one is still available for use.
• Jarno: came to mind: with XSLT 2, if open doc for reading, can have people hook in resolver. But what to do with result doc? Saxon has extension there we could use, but not part of Trax Java transformations, but could hook in. May get Saxon license to start developing more extensions.

Item 3: Upcoming 1.8 release

• Release Candidate coming shortly; final build near end of month.
• Updates coming to XSD's (fix to OASIS items) - should get those into 1.8. The pull request got mixed in with some other stuff so may be tricky to get them out. Can cherry-pick the changes, not as easy but can do it.

Item 4: Documentation changes / needs, and any updates due to coming changes

• Kris has some cleanup items to get in for Dev Ref section. Also have a plug-in for better looking PDF. Can get those in by end of this week.
• Kris now owns Don Day's old domain, ditaopentoolkit.org. Have not yet updated to store anything there - will for now use it as redirect to main OT page - but long run, can use that as home for documentation, gives us more freedom / power for better function. Kris has purchased 2 years of hosting at inmotion, seems nice friendly company that contributes to Open Source projects.
• Are 1.8 release notes up to date? If so, can edit. Jarno: may not be, should check against issues resolved for 1.8 milestone in github. Will try to update this week.

Item 5: Additional topics

• Code sprint: what's the stumbling block? Money for Jarno to get there? Think main issue is logistical - find a place to do it. Kris knows folks who may be able to help find a place in Munich. Also need some developers who are able to code and contribute. Not sure how many people at DITA Europe would have the skills to participate. George / Radu / Jarno could participate. DITA Europe is not that large, not as big as the US one, so hard to say - last year, about 100 people overall. Maybe Componize would participate? Bluestream, if coming? If get something organized, perhaps CIDM can also promote it when advertising the conference.
• Potential problem is that code sprints are generally most effective with people who know the code base; in our case it takes a while to get used to OT if you don't already know it. May not be that many people familiar enough with OT; most people who develop on top create custom plug-ins. Jarno - now working on preprocess code that nobody else seems to care about; maybe could get some people familiar with html / pdf that could contribute. Just worried there aren't enough people able to contribute in a day. George: so maybe more interesting idea to organize pre-conference event, full day of presentations on dita-ot internals? Maybe some vendors interested in sending to this event, also end users who want to understand?
• Robert wonders - in Providence, CIDM had a long extended panel discussion of toolkit. Maybe suggest to them a slightly modified version - two distinct, back to back sessions - first is a more general about-the-toolkit, something in between Matt's session from Providence and the panel discussion; second is a deeper technical dive, written up to try and draw in vendors, customizers, anybody who wants to dig into the code or logic. Had at least a couple of folks who expected that deeper dive from the Providence panel. Robert will send some notes to the group to make sure discussion continues.