incubator-general mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Kasper Sørensen <kasper.soren...@HumanInference.com>
Subject Potential proposal + request volunteers for champion & mentor roles - MetaModel
Date Mon, 28 Jan 2013 10:07:30 GMT
Dear all,

After some time spent (yeah, sorry quite a lot more time than we had hoped), we finally now
have the green light to officially post a proposal to the Incubator to take on MetaModel as
a project. Since it's been a while, I wish to call out again for sponsors. We found a few
some months ago, but we where out of contact for so long I feel it's most fair to ask again.

Please find below the draft proposal which we would love to further work on together with
sponsors before finally submitting it. (I tried posting it on the wiki but was not authorized,
probably because my account is a new one).

Best regards,
Kasper Sørensen

-------------------------------------------------

MetaModel - uniform data access across datastores
Proposal for Apache Incubator

-----------------------------

Abstract

MetaModel is a data access framework, providing a common interface for exploration and querying
of different types of datastores.

--------

Proposal

MetaModel provides a uniform meta-model for exploring and querying the structure of datastores,
covering but not limited to relational databases. The scope of the project is to stay domain-agnostic,
so the meta-model will be concerned with schemas, tables, columns, rows, relationships etc.

On top of this meta-model a rich querying API is provided which resembles SQL, but built using
compiler-checked Java language constructs. For datastores that do not have a native SQL-compatible
query engine, the MetaModel project also includes an abstract Java-based query engine implementation
which individual datastore-modules can adapt to fit the concrete datastore.

Background

The MetaModel project was initially developed by eobject.dk to service the DataCleaner application.
The main requirement was to perform data querying and modification operations on a wide range
of quite different datastores. Furthermore a programmatic query model was needed in order
to allow different components to influence the query plan.

In 2009, Human Inference acquired the eobjects projects including MetaModel. Since then MetaModel
has been put to extensive use in the Human Inference products. The open source nature of the
project was reinforced, leading to a significant growth in the community.

MetaModel has successfully been used in a number of other open source projects as well as
mission critical commercial software from Human Inference.

Rationale

Different types of datastores have different characteristics, which always lead to the interfaces
for these being different from one another. Standards like JDBC and the SQL language attempt
to standardize data access, but for some datastore types like flat files, spreadsheets, NoSQL
databases and more, such standards are not even implementable.

Specialization in interfaces obviously has merit for optimized usage, but for integration
tools, batch applications and or generic data modification tools, this myriad of specialized
interfaces is a big pain. Furthermore, being able to query every datastore with a basic set
of SQL-like features can be a great productivity boost for a wide range of applications.

Initial goals

MetaModel is already a stable project, so initial goals are more oriented towards an adaption
to the Apache ecosystem than about functional changes. We are constantly adding more datastore
types to the portfolio (currently a Salesforce.com web service consumer datastore is being
tested), but the core modules have not had drastic changes for some time. Our focus will be
on making ties with other Apache projects (such as POI, Gora and CouchDB) and potentially
renaming the 'MetaModel' project to something more rememberable.

--------------

Current status

Meritocracy

We intend to do everything we can to encourage a meritocracy in the development of MetaModel.
Currently most important development and design decisions have been made at Human Inference,
but with an open window for anyone to participate on mailing lists and discussion forums.
We believe that the approach going forward should be more encouraging by sharing all the design
ideas and discussions in the open, not only just the topics that have been "dragged" into
the open by third parties.  We believe that meritocracy will be further stimulated by granting
the control of the project to an independent committee.

Community

The community around MetaModel already exists, but we believe it will grow substantially by
becoming an Apache project. With MetaModel used in a wide range of both open and closed source
application, both at Human Inference (HIquality MDM), it's open source projects DataCleaner,
SassyReader and AnalyzerBeans and by other parties (such as the Quipo data warehouse automation
project), we believe that the critical mass to sustain a community is there.

Core developers

MetaModel was founded by Kasper Sørensen in 2009. Later it was incorporated as a core library
by Human Inference, meaning that more than 20 developers have been involved in its making
in this commercial setting. Furthermore a smaller number of contributors have submitted patches
for the library. Others have started building other interesting data-oriented libraries on
top of MetaModel, for instance the 'vasc' open source project by Willem Cazander, which is
an implementation of the Java Persistence API (JPA) for all the datastores supported by MetaModel.

Alignment

MetaModel already makes good usage of existing Apache projects such as POI, CouchDB and OpenOffice.
Furthermore developers from the Apache Gora project have indicated a need for a project like
MetaModel to solve specific uniform datastore access needs.

-----------

Known risks

Orphaned products

The contributors and the contributing organization (Human Inference) have a very strong dependence
on MetaModel already and will continue to have that for a long time. The continued need for
this vendor to support new types of datastores and maintain existing functionality will ensure
that MetaModel is not in the risk of being orphaned.

Inexperience with Open Source

MetaModel is already open source, and has been so for many years. Main contributors of the
project have also contributed to other open source projects such as DataCleaner and Apache
Mahout. The openness of Apache is arguably more extensive, but we are only encouraged and
delighted to be handling the project in a more open manner.

Homogenous Developers

Frequent committers are currently located in Denmark, The Netherlands and India. They are
used to working in a distributed environment.

Reliance on Salaried Developers

Most of the developers are paid by their employer to contribute to this project, but given
the dependence on MetaModel from both commercial and open source projects, the project would
continue without issue if no salaried developers contributed to the project.

Relationship with Other Apache Products

MetaModel depends on several Apache products including commons-lang, commons-io, commons-codec,
http-components, POI, CouchDB, OpenOffice and XMLBeans.
Furthermore MetaModel is built by Apache Maven.

An Excessive Fascination with the Apache Brand

The ASF has a strong brand, and that brand is in itself attractive. However, the developers
of MetaModel have been quite successful on their own and could continue on that path with
no problems at all. We are interested in joining the ASF in order to increase our contacts
and visibility in the open source world. Furthermore, we have been enthusiastic users of Apache,
and would feel honored by getting the opportunity to join the club.

-------------

Documentation

Information on MetaModel can be found at: http://metamodel.eobjects.org

Initial source

MetaModel has been developed since 2009 and have undergone a couple of major changes (indicated
by the 2.x and 3.x versions). The code is used in mission critical systems and is considered
very stable and high performing.

The source includes a fork of the xBaseJ project's code, which will be removed upon incubation.
This code was originally GPL licensed, but granted with a special license to MetaModel to
be forked and relicensed using the current LPGL license of MetaModel. Removal of the xBaseJ
code will effectively mean that the Apache variant of MetaModel will not have support for
dBase database files. We imagine that the dBase module could live on as a separate pluggable
module under the LGPL license, outside of Apache.

External dependencies

The dependencies all have Apache compatible licenses. These include BSD and MIT licensed dependencies.

------------------

Required resources

Mailing lists

The default set of mailing lists for an Apache project is considered sufficient for MetaModel.

Subversion directory

A subversion or git repository is needed. Currently MetaModel's code is hosted at http://eobjects.org/svn/MetaModel
but will be moved to an Apache repository.

Issue tracking

JIRA MetaModel (METAMODEL)

Other resources

If possible, a set of database servers (specifically MongoDB, CouchDB, MySQL, PostgreSQL,
MS SQL Server (Express), Firebird) should be made available for integration testing. Currently
this is done internally at Human Inference.

Initial committers

Kasper Sørensen (i.am.kasper.sorensen@gmail.com)
TODO

Affiliations

Kasper Sørensen - works at Human Inference
TODO

-------

Sponsors

Champion
TODO

Nominated mentors
TODO

Sponsoring entity
TODO

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message