Since writing the XmlBeans C++ proposal a few weeks ago, we have done more analysis of our planned approach, and we have come to the conclusion that there are significant and important differences between the XmlBeans architecture and that of our existing product that we referenced in the proposal as a starting point for the Apache project. In addition, we have since made progress on the XmlBeans C++ implementation with developers actively working on the project internally.
Our goal is clear, in that we plan to develop XmlBeans C++ to be compatible with the Java version, both architecturally and keeping to the API as closely as possible. There are many components in our existing LEIF product that will provide a starting point for XmlBeans C++ to jump-start the project. However, given the differences, we now plan to have our developers borrow/modify from this applicable code and then submit it to the Apache project during implementation. This approach will be a lot less confusing to the community, giving our developers the responsibility of incorporating any of our existing product code into the Apache project, avoiding that burden for other contributors. We can transfer our work under the auspices of Apache as soon as the project is approved and the environment is set up. Heidi. -------------------- Heidi Buelow Rogue Wave, a division of Quovadx > -----Original Message----- > From: Heidi Buelow > Sent: Sunday, May 15, 2005 11:12 AM > To: 'dev@xmlbeans.apache.org' > Subject: Proposal for xmlbeanscxx > > This is a proposal for an xmlbeans subproject which should be discussed > here > and voted on at the xmlbeans PMC before going to the incubator for a vote. > > Proposal to create a C++ version of the Apache XMLBeans project > > Submission date: 15 May 2005, Tim Triemstra, Heidi Buelow (TimT @ > RogueWave > dot-com, Buelow @ RogueWave dot-com) > > (0) rationale > > Apache currently maintains the XMLBeans/Java project, a tool that uses XML > Schema as a basis for generating Java classes used to easily access XML > instance data. The goal of the XMLBeans/C++ project is to provide a very > similar user experience for C++ developers as that provided by the > XMLBeans/Java version. This includes generating strongly-typed C++ classes > for the XML binding as well as providing low-level C++ APIs enabling > access > to the raw, underlying XML content. Where possible, the XMLBeans/C++ > project > will provide a nearly identical API and parallel architecture to that > provided by the Java counterpart, and will closely mimic both features and > schedule. > > Rogue Wave Software will help jump start this project by contributing the > existing XML to C++ binding code distributed as part of the LEIF product > to > the Apache community. This product has been in production for over two > years, and will provide a solid foundation for a C++ version of XMLBeans. > > (0.1) criteria > > Meritocracy: The C++ version of the project should adhere to the same > open, > merit-based community standards as other Apache projects. > > Contributions and Core Developers: Code being contributed is actually a > subset of the LEIF product. Currently, work is underway to remove any > proprietary dependencies from within the code. A "cleaned" version that is > ready for contribution will be made available to the community around the > middle of May 2005. This initial contribution will be fully functioning, > but > missing a few key features of the Java version of XMLBeans. > > The community should be able to immediately jump in and begin work, along > side dedicated Rogue Wave personnel, to turn the initial contribution into > a > feature-compatible version of XMLBeans for C++. > > Community: The community around the Java version of XMLBeans is already > vibrant and growing. There is every reason to believe many of those within > the existing community have experience and/or general interest in a > successful, compatible C++ implementation of the toolset. This should > ensure > an immediately active and vocal community, even if the primary interest is > in ensuring a similar experience between versions. There is already a > large > group of developers "in the wild" using the original XML to C++ binding > tool > that is serving as the original contribution for this project. These users > will be migrating to the XMLBeans/C++ code over time and have already > indicated a keen interest in tracking, and even participating in the > Apache-led project. > > Apache Alignment: The XML binding tool from which the XMLBeans/C++ port > will > morph does not include a full XML parser, but instead uses a small, > focused > parser designed expressly for tight C++ to XML binding support. The > general > purpose XML instance data APIs within XMLBeans will require additional > code > and may be a good fit for the Xerces parser to enable some required > features. > > Additionally, at least one member of an existing Apache project has > expressed interest in a project like XMLBeans/C++ to aid in his current > work. Higher-level projects such as Axis for C++ seem to be natural > beneficiaries of this work. > > (0.2) known risks > > Orphaned Products: One of the first questions when a commercial entity > offers code to the public is "will this code be abandoned?" To be clear, > Rogue Wave decided to initiate this process due to its own significant > interest in the XMLBeans feature set already present in Java, and its need > for similar features in C++. We have significant product goals moving > forward around XML for which the XMLBeans model fits perfectly. Rather > than > simply copy the design and keep the project closed, we felt that the > downstream products would reap significant benefits from opening the XML > layer to the community. Keeping the C++ version in sync with the Java > version, even if it were to not be accepted as an Apache project, can only > help its overall adoption. > > Additionally, a diverse committer base is a strong goal for this project. > Numerous users of the contributed XML to C++ binding tool have indicated > interest in participating, many of whom have critical projects dependant > on > this work and resources available to continue to ensure the viability of > the > project well into the future. > > Inexperience with Open Source: Like many companies entering this arena, we > have limited experience working on open source projects. Our primary goal > is > to foster an active community around XMLBeans/C++ so advice will be taken > to > heart, and significant resources will be dedicated to the project to get > it > off the ground. However, we hope that our experience working in open > standards groups will aid in the transition to the open source community. > > Initial Reliance on Salaried Workers: Due to the rapid process of > transitioning to work on the open source XMLBeans/C++ project, only a few > of > the listed contributors will be outside the commercial realm. We expect > this > list of external volunteers to grow significantly after the initial public > code drop. > > Licensing, Patents, Miscellaneous Legal: We are conducting a legal review > of > the code and existing contracts. This review should be done shortly and > any > code contributed will be licensed under the latest ASF terms. > > Commercial Interest: XMLBeans/C++ will be maintained as an open source > Apache project, with all relevant enhancements contributed to the > community. > Additionally, there is every intention to use XMLBeans/C++ within future > commercial products, thereby resulting in even greater testing and user > exposure. It is expected that other companies may well wish to use the > project's code within their own commercial endeavors, which of course > would > be fine. > > (1) scope of the subproject > > The XMLBeans/C++ subproject will conform to the identical scope as that > laid > out for the partner Java project. Special care will be taken to implement > features and add conveniences that would be expected by a C++ developer. > > For clarity, the goals of the XMLBeans/C++ project are: > > Generation of plain C++ classes to model XML Schema Validation of C++ > objects against the source XML Schema Access to partial document instance > data (fragments) Efficient "parse as necessary" access that forgives extra > data Access to the full XML infoset > > > > (2) identify the initial source for subproject code > > Some background information may be found on the LEIF product and the > associated Data Tier. > > LEIF product page (http://www.roguewave.com/products/leif) > > The C++/XML binding contribution code can be found at the following link: > > xmlBeans open source information > (http://www.roguewave.com/opensource/XMLbeans.cfm) > > (3) identify the ASF resources to be created > > (3.1) mailing list(s) > > xmlbeanscxx-dev > xmlbeanscxx-user > xmlbeanscxx-commits > > (3.2) SVN repositories > > xml-xmlbeanscxx > > (3.3) Bugzilla > > xml xmlbeanscxx > > (4.0) identify the initial set of committers > > This is a preliminary list that will be updated with volunteer members. > > Tim Triemstra (TimT @ RogueWave dot-com) > John Hinke (Hinke @ RogueWave dot-com) > Heidi Buelow (Buelow @ RogueWave dot-com) > Allen Brookes (ABrookes @ RogueWave dot-com) > David Haney (David.Haney @ RogueWave dot-com) > Michael Yoder (Michael.Yoder @ RogueWave dot-com) > > (5) identify apache sponsoring individual > > Cliff Schmidt, of the XMLBeans/Java project, has volunteered to sponsor > this > project. > > Cliff Schmidt (CliffS @ Apache dot-org) > > (6) open issues for discussion > > The original code contribution has a lot of proven code for creating a > binding between XML Schema and C++ classes. However, the contribution will > require a significant overhaul, and even complete re-writes in some areas, > in order to reach compatibility with the XMLBeans/Java version. Detailed > differences will be discussed openly within the community so that an > appropriate plan for each area can be reached. This proposal is not the > best > place to lay out all the technical details, however you will find below a > preliminary list of issues, priorities, and estimated timelines. > > API discrepencies between contributed code and XMLBeans 2.0 The > contributed > code includes a very similar type-binding API in C++ to that provided in > XMLBeans for Java. However, the XMLBeans cursor access to the underlying > infoset has no equivilant in the contributed code and will require > extensive > work to the underlying object model to support. > > Addressing this deficiency is considered the highest priority task facing > the C++ project. > > Several options are available including porting the XMLBeans object model > completely over to C++ or looking for ways to augment the contributed > code. > Most likely a near-complete rewrite of this section will be the best > approach. > > Rough estimates for a working implementation of the cursor feature are > between three and five months Code generator > > The contributed C++ code is based on a proprietary code generator, > although > it too is written in Java (not C++). The long-term plan is to consolidate > both C++ and Java projects under the same code generator, most likely the > one already included in the Java version of the XMLBeans. > > This is not considered a priority-1 task at this point, more of a long > term > goal and convenience. If it saves time in other priority tasks it will be > looked at closely, however. > > Timeline for this transition is 5-7 months, but may be very much sooner > depending on available resources. > > Platform support > Unlike Java, the C++ port will have significant issues around platform and > compiler support. In order to help with this the project will be using the > C++ standard library, and may in fact benefit greatly from the proposed > Apache C++ StdLib (STDCXX). Initially the native StdLib will be used. > > Good platform support will be very important so even the early builds > should > be done on Windows and Linux, and preferably at least one Unix variety. > Long > term the matrix should be very large. > > The initial contribution is working on Linux but it is really just a > preview > look at code, not a true product build. Windows and Linux support should > accompany the initial beta release, with Solaris, HP-UX, AIX, and other > Linux distros following within three months. > > Estimated timeline (very rough) > Initial submission of proposal and example code: 12 May 2005 Apache > structure in place and code building: 1 July 2005 Re-write of internal > object model building: 1 September 2005 XMLBeans/Java 2.0 feature > equivilant > beta: 1 December 2005 > > This timeline is based on a very preliminary discovery of issues and is > based on limited resource information. There is little chance for this > schedule to be elongated by very much time as there are already a few > dedicated resources assigned to the project, but there is also little > known > about how much a concerted community effort could speed things up. In > general the feeling is that a greater community involvement will greatly > improve the test suite, identified bugs, and platform support - not > necessarily the initial speed of feature development. > > > > -------------------- > Heidi Buelow > Rogue Wave, a division of Quovadx > > --------------------------------------------------------------------- > To unsubscribe, e-mail: [EMAIL PROTECTED] > For additional commands, e-mail: [EMAIL PROTECTED] --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
