This Bugzilla instance is a read-only archive of historic NetBeans bug reports. To report a bug in NetBeans please follow the project's instructions for reporting issues.
Summary: | MDR Architecture Review | ||
---|---|---|---|
Product: | java | Reporter: | Martin Matula <mmatula> |
Component: | Unsupported | Assignee: | Martin Matula <mmatula> |
Status: | RESOLVED FIXED | ||
Severity: | blocker | CC: | apireviews |
Priority: | P3 | Keywords: | API_REVIEW |
Version: | 3.x | ||
Hardware: | PC | ||
OS: | Windows ME/2000 | ||
Issue Type: | TASK | Exception Reporter: | |
Bug Depends on: | 41488, 41556, 41557, 41558, 41559, 41560, 41561, 41562, 41563, 41564, 41565, 42082 | ||
Bug Blocks: |
Description
Martin Matula
2004-02-29 22:35:39 UTC
The target milestone should be set to NB 4.0/PromoD - for some reason it does not appear in the list of target milestones for this module, so I am not able to formally set it. The review will be lead by Pavel Buzek and the inception review meeting shall be held on Mar31, 2004, if possible. Other reviewers include Tomas Pavek, Miloslav Metelka and me. here are my initial comments that I would like to discus in the meeting: I'd suggest to separate the API needed for module writters who need to: 1. work with a meta model provided by somebeody else 2. provide support for a new metamodel 3. provide a new implementation of any part of MDR This would make the whole documentation more user/task oriented. BTW: This is what I would also expect to see in Use Cases document. As for the 3rd category: Why id the MDR API a public contract and not a private contract between your modules? Who is expected to install a new MDR repository implementation? Same question about XMIReader, maybe JMIMapper, etc. Is Persistence SPI private? From the architecture document it seems like it is a contract but I have not seen it in answers or javadoc (I would say it should be private, I am just checking). As for recomending org.netbeans.jmi.* package for generated interfaces: the convention is to use o.n.modules.* or o.n.api.* or o.n.spi.* Why cannot it be o.n.m.jmi or o.n.api.jmi? suggested TCA: deploy-packages - what prevents you to do it before commit? suggested TCA: do i18n before commit perf-startup: I assume the inicialization is done lazilly. What is the performance of first call in MDR? perf-exit: What is the performance of the tasks that are done on IDE shutdown? perf-limit: What is the size of repository on disk? How much overhead does it have compared to size of stored data? perf-mem: What is a practical size of this cache for real-life examples (such as Java refactoring) and how does it impact performance? perf-progress: How long does it take to boot the repository? Is this also done when upgrading to newer version of MDR? Opinion document posted: http://openide.netbeans.org/tutorial/reviews/opinions_40618.html assigning back to Martin to work on the issues and prepare for the final review. Thanks! done, I guess. |