Stefane Fermigier

New Nuxeo Architecture Slides Published

I’ve just posted on slideshare a set of slides that have been used at recent customers and partners presentations about the Nuxeo architecture.

| View | Upload your own

Hope you’ll enjoy it. There are some more in the works.

Nuxeo EP 5.1.3.2 released

We have released Nuxeo EP 5.1.3.2 earlier this week. This is a maintenance release primarily focussed on bug fixes and small improvements.

You can download it as

a multi-platform installer

or:

a simple ZIP for headless installation

The changelog for this release is available.

Minor, but anyway noteworthy improvements include:

Nuxeo EP now works on a Java 6 JVM, which can lead to significant performance improvements (up to 100% for certain workloads, according to some internal benchmarks).
Full text indexing has been improved and made faster.
Some tweaks have been done to enable communication with the new (soon to be released and announced) “LiveEdit” plugin for Internet Explorer, Firefox and MS-Word.

Interview on ComputerWorld UK

I've been interviewed by Glyn Moody, one of the best journalists covering open source, for the ComputerWorld UK web site.

France is not a country many would associate with free software startups, but that's changing – not least because the French government is showing itself far more receptive to open source than its UK counterpart. One of the leading companies of this new Nouvelle Vague is Nuxeo, which was set up by Stefane Fermigier, now its CEO.

Link: Open Enterprise Interview: Stefane Fermigier.

Nuxeo EP 5.1.3 and 5.1.3.1 released

Nuxeo EP 5.1.3 has been released in January, with 235 enhancements over the previous version, Nuxeo EP 5.1.2, which had been released in October.

A bugfix release (5.1.3.1) was also done a few weeks after with 20 bugfixes and enhancements.

Download it now (107 Mb).

Upcoming Nuxeo 5.1.3 release + Updated roadmap

The Nuxeo EP 5.1.3 release is coming along nicely after a small delay due to the holiday break, and will be tagged in a couple of days.

The roadmap for the project in 2008 has been updated accordingly.

Feel free to discuss it in the mailing list or the forum if you have questions, suggestions or to supply missing information.

As can be seen, Nuxeo 5.1.3 will not be just a maintenance release, as we’ve been able to add new features by creating new plugins (ex: WebDAV, Portlets, SSO, etc.) thanks to a now quite stable infrastructure and API.

Nuxeo 5.1.3 will also be based on Nuxeo Runtime and Core 1.4, which feature some improvements while staying compatible with the previous version (1.3). The switch to Nuxeo Core 1.4 was done in December and has proved very stable.

After the 5.1.3 release, we’re going to focus on the 5.2 (trunk) work, so as to move to Seam 2 for the web platform. Most new features will still be developed as new components, and will either be delivered with the next maintenance release (Nuxeo 5.1.4) or with Nuxeo 5.2, or both, depending on technical feasibility and customers needs or community contributions.

Nuxeo 5.1 RC released - GA release scheduled for next week

We have been so busy the last couple of months working on customers projects that the 5.1 release has slipped a bit, but I’m happy to report that we have just released Nuxeo 5.1 RC.

The final release (Nuxeo 5.1.0.GA) will be made next week, and then we’ll spend the rest of August:

finishing and polishing the Nuxeo Book.
planning and starting the next iteration of Nuxeo (5.2, cf. the current roadmap).
Working on new customers projects (including some that feature the Apogee Project which has already seen a recent surge of activity).
getting some holidays :)

Interview rat�e du Debian Project Leader dans Le Monde

Le quotidien Le Monde a r�cemment fait para�tre une interview du “Debian Project Leader”, Samuel Hocevar. La lecture de cette interview m’a passablement agac�, car d’une part elle propage � mon sens un certain nombre d’id�es re�ues sur le libre (“logiciels d’informaticiens pour informaticiens”) plut�t que d’en faire la promotion aupr�s d’un public assez large, et d’autre part elle s’attache plus � faire la promotion de Wikipedia que du logiciel libre.

Plut�t que de critiquer point par point les r�ponses de Samuel aux questions du journalistes, j’ai pr�f�r� refaire mes propres r�ponses � ces m�me questions.

Que pensez-vous du succ�s grandissant des logiciels Firefox ou OpenOffice ?

Le navigateur web Firefox et la suite bureautique OpenOffice.org partagent un certain nombre de caract�ristiques qui expliquent leur succ�s actuel, aupr�s du grand public et de certaines administrations ou entreprises:

Ce sont des logiciels qui r�pondent aux deux principaux besoins g�n�riques des utilisateurs d’informatique: l’acc�s au Web et la bureautique.
Ce sont des logiciels multi-plateformes: ils tournent � la fois sous Windows (qui reste la plateforme dominante du march�), sous Mac OS et sous Linux.
Ce sont des projets matures: Firefox est issu de la base de code du navigateur Netscape d�velopp�e dans les ann�es 90, OpenOffice.org de la suite StarOffice dont le d�veloppement a d�marr� en 1994.
De plus, il s’agit de projets qui disposent d’une force de travail importante, constitu�e en partie de personnel d’acteurs majeurs de l’informatique (IBM, Google, Sun, Novell…) qui ont un int�r�t strat�gique � contrer l’h�g�monie de Microsoft sur le poste de travail.
Il y a un lien tr�s fort entre ces deux logiciels et les standards ouverts sous-jacents: les standards du Web pour Firefox, le standard ISO Open Document Format pour OpenOffice.org. La lutte de lobbying intense � laquelle Microsoft se livre depuis plus d’un an pour faire normaliser son propre standard “Open XML” aupr�s de l’ISO (en d�pit du bon sens: pourquoi cr�er une deuxi�me norme alors qu’il en existe d�j� une?) et selon des m�thodes peu consensuelles montre l’importance strat�gique des standards dans l’informatique actuelle.

Il faut noter cependant une diff�rence importante: le principal concurrent de Firefox est Internet Explorer (et, dans une moindre mesure, sur plateformes Mac OS, Safari) qui est int�gr� comme navigateur par d�faut dans les syst�mes Windows de Microsoft, donc vu comme gratuit par les utilisateurs, alors qu’OpenOffice.org se positionne principalement face � la suite Office de Microsoft, qui est on�reuse:

Pour arriver � convaincre une proportion significative d’utilisateurs d’installer Firefox plut�t que le navigateur par d�faut, les d�veloppeurs de Mozilla doivent se diff�rencier par la qualit� et les fonctionnalit�s de leur produit. Ainsi, face � Microsoft qui, une fois qu’il a cru avoir gagn� la “guerre des navigateurs” face � Netscape au d�but des ann�es 2000, a cess� toute innovation sur son navigateur, Mozilla a introduit des dizaines d’innovations comme par exemple la navigation par onglets ou les bloqueurs de popups, innovations pl�biscit�es par les utilisateurs � tel point que Microsoft a �t� oblig� de les copier dans IE 7.
Dans le cas d’OpenOffice.org, la diff�renciation se fait le plus souvent par le prix: la suite Office 2007 de Microsoft co�te, typiquement pour une PME, de 500 � 950 euros par poste, ce qui est du m�me ordre de grandeur que le prix d’un ordinateur bureautique d’entr�e voire de milieu de gamme. C’est un co�t important qui rend tentante l’offre gratuite d’OpenOffice.org.

Justement, comment voyez-vous l’avenir des logiciels libres grand public ?

Le succ�s de logiciels comme Firefox, OpenOffice.org ou VLC (logiciel de lecture vid�o lui aussi multi-plateformes) montrent que des logiciels clefs, qui peuvent repr�senter jusqu’� 80 ou 90% de l’utilisation quotidienne de l’informatique par un grand nombre d’utilisateurs, peuvent �tre des logiciels libres.

Ce succ�s d’un petit nombre de logiciels g�n�ralistes, et d’une “longue tra�ne” de logiciels plus sp�cialis�s, sur des plateformes comme Windows et Mac OS, permet d’�duquer le grand public sur l’existence et la qualit� des logiciels libres, et peut en amener un certain nombre � vouloir �galement s’int�resser � Linux en tant que syst�me d’exploitation pour postes de travail. Les principaux �diteurs de syst�mes d’exploitations bas�s sur Linux - Red Hat, Novell, Mandriva, Ubuntu - constatent actuellement une demande du grand public, et de la grande distribution, sur ce secteur, et plusieurs indices, dont le “flop” du lancement de Vista, laissent � penser que 2008 sera l’ann�e du d�collage de Linux aupr�s d’une partie du grand public.

Signalons par ailleurs un autre facteur de diffusion des logiciels libre: le logiciels embarqu�s dans du mat�riel sp�cialis�. En France, la Freebox, la NeufBox et de nombreuses autres “appliances” int�grent d�ja depuis plusieurs ann�es un noyau Linux et de nombreux logiciels libres. Dans les pays �mergents, des initiatives comme le One Laptop Per Child sont �galement une fa�on de diffuser des logiciels libres aupr�s de millions d’utisateurs, qui ne connaissent pas Windows et qui sont donc vierges de tout a priori.

De plus en plus d’entreprises priv�es et d’administrations passent aux logiciels libres, d�sormais consid�r�s comme des “concurrents” par Microsoft. Existe-t-il encore des freins � leur g�n�ralisation ?

Dans le domaine des logiciels d’entreprises, la donne est tr�s diff�rentes: il n’y a pas un seul acteur qui domine outrageusement le march�, mais plusieurs acteurs dominants: Microsoft certes, mais aussi SAP, IBM, Oracle, EMC, etc.

Sur ce secteur, et principalement dans le domaine des logiciels serveurs, d’abord au niveau des logiciels d’infrastructure (communication, bases de donn�es, monitoring, etc.) et progressivement au niveau des logiciels applicatifs (CRM, ERP, GED, ECM, etc.) les logiciels libre se sont dans certains cas d�ja impos�s (ex: Apache) et dans la plupart des autres, connaissent une progression rapide.

Un frein notable � cette progression est l’ensemble des pratiques anti-concurrentielles de Microsoft, d�nonc�es par l’ensemble de l’industrie, et qui lui ont valu de tr�s nombreux proc�s et un certain nombre de condamnations.

A part cela, il est plus appropri� de parler d’inertie que de freins, car il s’agit de faire changer les mentalit�s des d�cisionnaires, de former les informaticiens de terrain, et de remplacer des investissements qui s’amortissent sur de 5 � 10 ans.

Une chose est claire pour toute le monde cependant: le logiciel libre a chang� la fa�on dont les logiciels sont d�velopp�s par les �diteurs. Ceux-ci se reposent de plus en plus (31% en 2006, probablement plus de 50% en 2007 ou 2008) sur des “briques” logicielles libres, mais aussi “ouvrent” de plus en plus leur mod�les de d�veloppement. D’ici 5 ans, je suis certain que la plupart des �diteurs de logiciels d’entreprises auront int�gr�, d’une fa�on ou d’une autre, une partie des m�thodes du libre dans leurs d�veloppement.

Comment voyez-vous l’�volution de la distribution de contenus culturels ?

On est dans un cas d’�cole d’innovation disruptive, selon le mod�le de Clayton Christensen:

Les anciens acteurs dominants, les “majors”, s’accrochent � leur ancien mod�le � pr�sent obsol�te et militent pour des l�gislation drastiques dans ce domaine.
Une partie du public � trop rapidement adopt� les pratiques d’�change en pair � pair ou sur des sites de mise en ligne de contenus, sans int�grer les limites l�gales et morales de leurs pratiques: �changer des contenus libre et contribuer de mani�re communautaire � leur d�veloppement, c’est bien; �changer des contenus soumis � des restrictions d’usage en dehors du droit � la copie priv�e, c’est mal.
Les pouvoir publics, faute d’une vision claire sur ce dossier, et sous l’influence des lobbies du pass�, ont vot� des lois tr�s dures (DADVSI) visant � maintenir le statu quo.
Un petit nombre d’acteurs, � commencer par Apple, a saisi au moment opportun la rupture technologique et s’en est servi pour reconfigurer la cha�ne de valeur de la diffusion des contenus autour de leur offre de produits (ex: l’iPod) et de services (ex: iTunes), r�alisant ainsi le “hold-up du si�cle” sur l’industrie musicale, et bient�t sur l’industrie cin�matographique.

A mon sens, l’un des r�le majeurs d’un projet comme Wikipedia, qui montre la force mais aussi les limites du mod�le de d�veloppement coop�ratif, est d’�duquer le public sur ces questions de production communautaire de contenu et sur les questions de droits d’usage des contenus, de montrer qu’il y a plusieurs mod�les possibles, chacun avec ses forces, ses faiblesses et ses tabous, et qu’il est important de les conna�tre pour se comporter de mani�re citoyenne.

JavaOne Nuxeo Slides: now in original format

The slides of our presentation at JavaOne have now been published by Sun (minus an URL error from this page which I hope they will correct soon).

Now that Sun has made them public, I have uploaded them to SlideShare, where they are available alongside other presentations tagged “javaone”.

Are Nuxeo's open source projects truly community-driven? You bet they are!

Dion Almaer wrote yesterday a blog entry with a list of criteria that, in his opinion, make an open source project truly community-driven (which is, of course, a Good Thing(tm)).

I had some apprehension, while reading the blog entry’s title, because, of course, we want to be community-driven (that’s one of the criteria BTW), but what if we had forgotten something important?

Fortunately, the short answer is “of course we are community-driven”. With Dion’s criteria, I can confidently self-grade us at A+ (or 20/20, for french-educated people). Here are the criteria and my comments:

If you don’t have any commiters from outside of your company. You probably aren’t community driven.

Half of the commiters in the Nuxeo project are not Nuxeo employees, and we are actively trying to recruit more contributors.

If you didn’t spend time cleaning up documentation for the community when you opened it up. You probably aren’t community driven.

We’ve produced already several hundreds of pages of users and developers documentation, and we’ve gone through great efforts in producing this documentation in time for the first public releases of the software.

If your users haven’t helped with the documentation if it is lacking. You probably aren’t community driven.

We got great feedback from our users regarding the documentation, and we’ve already had user-contributed sections in some of our manuals.

If you do not have some kind of forums/lists where people help each other out. You probably aren’t community driven.

We have several mailing lists for all of our projects, with a total of about 1500 subscribers.

If you aren’t willing to put in a lot of effort to build your community to get true benefits. You probably aren’t community driven.

Yes, we’re putting a lot of efforts, and we get the rewards of course.

But there is something more important, that is not explicitly stated in Dion’s criteria: we have designed the Nuxeo software with the explicit goal of creating an Architecture of Participation (a term coined by Tim O’Reilly, Dion’s boss, BTW). Our creation of Nuxeo Runtime, the OSGi-based plugin system (inspired by Eclipse’s), our use of a component framework like JBoss Seam for our webapp, are consequences of this vision, which comes from years of experience working with system integrators and ISV.

Extensibility at all levels is one of the majors criteria of a successful historical open source project (think Eclipse’s or Mozilla’s plugins, Emacs scripts, LaTeX packages, etc.). Monolithic apps usually don’t make great open source communities. That’s a lesson we will never forget!

Back from BrainCamp Ouverture 2007

As some people know already, I have participated in the creation of Ouverture (n� “Open Source Valley”), the Competitiveness cluster dedicated to Free and Open Source Software in the Paris area.

We are still waiting for the official stamp of the French Government, which will start the operational life of the cluster by bringing Government funding to R&D projects carried out by the cluster’s members: public research labs, universities and companies. Since there seems to be some, let’s say “bureaucratic delays”, we had decided to start some activities that don’t depend on the official stamp, namely bringing together all the cluster’s participants to brainstorm ideas for projects that would be later submitted to the cluster.

This first event was christened “BrainCamp 2007” and happened today at ENSTA, a Parisian engineering school that has been a pionneer in the use of open source software in higher education in France.

Like every SME that participated in this BrainCamp, I had 5 minutes to present Nuxeo to the other participants: about 15 research labs, 20 SMEs (including the biggest names in the French open source market: Mandriva, Idealx, Linagora, XWiki, etc.) and 4 big companies (which happen, can you believe it, to be all partners of Nuxeo: Capgemini, Unilog, CS, Sun).

My point was to present the most innovative aspects of the Nuxeo 5 project: its use of Nuxeo Runtime to develop modular applications that target both application servers and rich clients, the Nuxeo Core embeddable document management core, and the Nuxeo EP and Nuxeo RCP platforms. Hence we are looking for partners that would like to work with us on all kind of levels:

Nuxeo Runtime: either developers of other kinds of enterprise software (ERP, CRM, Enterprise Wikis…) that would want to benefit from the modularity and extensibility that it provides, or people who would want to work with us on targeting other deployment platforms than JBoss AS 4 and Eclipse (we’re specially looking for volunteers for helping with porting Nuxeo Runtime to JBoss 5, JOnAS 5, Geronimo, Glassfish, for instance).
Nuxeo Core: developers that who would like to extend to core with new kinds of document storages than the JCR, for instance a RDBMS (via Hibernate), an OODBMS (like db4o) or an XML database, add application-level replication / synchronization, etc.
Business components developers that would extend the platform with more specialized components.
Vertical application developers that would base their software on the Nuxeo platform (we already have a bunch of companies we are working with on this matter).
Other kinds of applications developers (ex: CRM, ERP, etc.) that would want either to embed Nuxeo as a document management core in their system (this makes completely sense for a system like a CRM that has to manage document besides more structured data) or just connect Nuxeo to their software.

For those who prefer slides, here they are:

Or as a PDF file, if you prefer.

If you are interested in collaborating with us, either contact me (sf@nuxeo.com), or join the mailing list.

JavaOne 2007: our talk about Nuxeo has been accepted

I’m very happy (some would say "ecstatic" ;) ) to report that our talk proposal for the JavaOne conference has been accepted by the program committee.

It will be called “Building an embeddable Enterprise Content Management core” and will be an in-depth overview of the Nuxeo 5 architecture, as proposed in our abstract:

This session describes the architecture and implementation of an embeddable, extensible Enterprise Content Management core for Java EE and simpler platforms.

The presentation starts by describing the general architectural concepts that are used as building blocks:

a schema and document model, reusing XMLSchemas and making good use of XML namespaces, where documents types are build using a number facets,

a repository model, using hierarchy and versioning, with JCR (JSR-170) being one of the possible backends,

a query model, based on JPQL (JSR-220) and reusing the path-based concepts from JCR,

a fine-grained security model, compatible with WebDAV concepts and designed to provide flexible security policies,

an event model using synchronous and asynchronous events, allowing bridging through JMS (or other systems) to other event-enabled frameworks,

a directory model, representing access to external data sources using the same concepts as for documents but taking advantage of the specificities of the data backends,

Suitable abstraction layers are put in place to provide the required level of flexibility. One of the main architectural tasks is to find commonalities in all the systems used (or whose use is planned in the future) to have a minimal number of concepts learned and used by the user of the framework. The result is a set of concepts that are fundamental to Enterprise Document Management and are usable through direct Java APIs, Java EE APIs or SOA. The presentation will show, for each of the main components, which challenges have been met and overcome when building a framework where all components are designed to be improved and replaced by different implementations in the future without sacrificing the backward compatibility with existing ones.

The described implementation, Nuxeo Core, can be embedded in a basic Java framework based on OSGi (like Eclipse), or in one based on Java EE, according to the needs of the application using it. This means that the core has to function without relying on Java EE services but also has to take advantage of them when they are available (providing clustering, messaging, caching, remoting, and advanced deployment).

The session includes a demo. Attendees should have intermediate knowledge of Java technology concepts and design patterns, and an understanding of the Content Management problem space.

Announcing Nuxeo Weekly News

I have just published the second issue of Nuxeo Weekly News, the newsletter we have started to publish every monday to keep the Nuxeo community informed of what’s going on with the project, and to provide some insight about where we are heading.

For those who missed it, the first issue is still available.

I will appreciate any feedback (positive or negative) about it.

Slides de la pr�sentation "Nuxeo Runtime" de Solution Linux 2007 en ligne

Je viens de mettre en ligne les slides de la pr�sentation de Bogdan sur Nuxeo Runtime.

La pr�sentation avait lieu lors de la session “J2EE” des conf�rences techniques du salon Solutions Linux 2007.

Les slides sont aussi disponibles en PDF.

JBoss World Berlin 2006 slides on SlideShare

I have recently published the slides for Nuxeo’s presentations at JBoss World Berlin on Slideshare,

First one, the business one, was called “Nuxeo - An Open Source ECM Vendor”, and is available here or just below:

The second one, the technical one, was called “Nuxeo EP 5 - A Seam Case Study”, it is also available just below:

Both presentation are also available more traditionnally as PDF here (technical one) and here (both of them)

Stangely, the technical presentation go 8 times as more hits that the business one ;)

Seam 1.1 almost there + insightful interview on InfoQ

As everyone should know now, Seam is an important part of our Nuxeo 5 ECM stack.

We are currently using Seam 1.0.1, but it is nice to learn that Seam 1.1 is coming along, and it’s going to be very good.

There is a very nice (like always, on InfoQ) interview of Gavin King, Seam’s creator, that explains in details where Seam is headed (rest assured that Nuxeo 5.1 will include Seam 1.1, but let’s focus on the Nuxeo 5.0 release now ;) ).

Java, OSS and innovation

“Cristiano”, from Faqs.it, has written a comment about our switch from Python to Java in his blog.

Here is my comment about his comment.

Hi Cristiano,

Thanks for your comments about our Java switch. I fully agree with you that a lot of innovations come from the OSS ecosystem, but there are some wrong facts in your remarks that I have to correct, including the following:

Python, for instance, did actually copy Java’s syntax (and concept) for annotation (called decorators in Python-speak), not the other way around.

Generics are irrelevant in dynamics languages, hence dynamic languages can’t have “out-innovated” Java on that matter.

The sentence “Python doesn’t fit in the enterprise-class web apps market” is in part irrelevant for us, because we are not only in the web apps market, but also in the rich client market (with Eclipse RCP).

Your opposition (in the last sentence) of “Java vs. OSS” is completely disconnected from the reality: Java is GPL now (or will be RSN), Java OSS communities (JBoss, Spring, Apache, Eclipse, Nuxeo…) are among the most active (if not the most active) OSS communities now. And some of these communities have a lot of money at their disposal, because these are projects for building enterprise-class OSS software (or middleware), backed by serious companies like Sun, IBM, Red Hat, or Nuxeo.

IMHO, OSS is now driving, in many ways, the development of Java, and this means that there is no need to oppose Java if you are (like I am) an OSS believer.

Seam 1.1 beta1 released

Gavin King has released last week the first beta of JBoss Seam 1.1, but hasn’t made any spectacular announcement for it, besides a post in the JBoss Seam forum.

According to his release notes, the major new features of Seam 1.1 are:

The Seam Application Framework for data-oriented applications
Asynchronous methods and events
Integration of Ajax4JSF for partial page refresh
New concurrency model for Ajax-based applications
Efficient clustering of Seam-managed extended persistence contexts and JavaBean components
Support for atomic conversations
Enhanced configuration via components.xml
Exception handling via annotations or exceptions.xml
Page parameters for RESTful applications
Page fragment caching via <s:cache/>
“Decoration” of invalid input via <s:decorate/>
Themes
Support for the Sun JSF 1.2 reference implementation

Michael Yuan, author of the upcoming Prentice Hall Seam Book, also has some comments on the new features.

As developers of the Nuxeo 5 open source ECM platform, which relies heavily on Seam for its web presentation layer, we are both excited by this new release, and curious about other projects (open source or not) that are using the Seam framework.

Maven workshop with Vincent Massol

Vincent Massol, of Maven and Cargo fame, came to visit the Nuxeo 5 team yesterday to give us some advice on how to best use Maven 2 to build and deploy Nuxeo 5 and ensure the best possible software quality for the product.

So, expect a few changes (= improvements) to the project’s SVN structure, the Nuxeo 5 Maven site and of of course the build system if you are already a developer (if you’re not, I strongly encourage you to consider becoming one.)

And, BTW, we have identified a few features we could add someday to Maven (as a plugin) or Cargo, so expect some contributions from us in the future (expect, but don't count on it before the Nuxeo 5.0 release next month.)

Another review of Mike Milinkovich's keynote

Well, it seems Mike doesn’t have enough time to prepare a new keynote each time he has a speaking engagement ;)

This review of his keynote at CSS2006 seems suspiciously similar to the one I saw in Esslingen am Neckar two weeks ago.

Anyway, just kidding: this was an awesome keynote and if you missed it, or my summary of it, or the CSS 2006 version, you still have my and Matt Raible’s blogs entries to read it (I wonder if the slides are available somewhere, though).

"Eclipse Now You Can" slides are now online

The slides for the presentations given at the "Eclipse Now You Can" (including mine of course), are now online.

The conference was a smashing success, btw, with 100 attendees for 100 seats ;)

I'm now an Eclipse pundit ;)

It's official now :) !

1. I have given a talk at an Eclipse conference.

2. My blog (like Eric's and Bogdan's is now aggregated on Planet Eclipse.

Nuxeo 5 ECM project passes the 40 kloc mark

The Nuxeo 5 code base, hosted on svn.nuxeo.org, has just reached the 40 kloc (= 40000 single lines of code, excluding comments and blank lines, as computed by the sloccount utility) tonight:

SLOC    Directory       SLOC-by-Language (Sorted)
9854    NXCore          java=9854
5211    NXRuntime       java=5211
4854    NXJCRConnector  java=4854
3430    NXThemesCore    java=3430
2301    Nuxeo           java=2245,jsp=56
1646    NXCoreFacade    java=1646
1356    NXCoreAPI       java=1356
1311    NXTransform     java=1311
1235    NXJBossRuntime  java=1235
1163    NXRelations     java=1163
937     NXThemesJsf     java=937
842     NXTransformPlugins java=842
741     NXMimeType      java=741
615     NXAction        java=615
476     NXPlatform      java=476
473     NXSQLDirectory  java=473
451     NXAudit         java=451
404     NXEvents        java=404
360     NXAdobeLiveCycle java=360
339     NXJobs          java=339
312     NXTypeManager   java=312
308     NXThemesJsfFilters java=308
302     NXServerCache   java=302
242     NXThemesJsfEditor java=242
207     NXContextInvalidationManager java=207
186     NXClientCache   java=186
178     NXJBossCache    java=178
169     NXDirectory     java=169
133     NXAuditClient   java=133
95      NXThemesFragments java=95

Slides for my Nuxeo RCP and Apogee talk at the Eclipse Summit Europe

I’m just coming back from the Eclipse Summit Europe 2006 which was a great conference (more reports on this blog later).

I had the chance to do a short (20 minutes) but exciting ( ;) ) presentation of two projects we are involved in: Apogee and Nuxeo RCP.

My slides are now online on nuxeo.org, or on the conference website.

Eclipse Summit report, day 1, Mike Milinkovich's keynote

Mike Milinkovich started the conference by highlighting the four ingredients of a great platform:

Great technology
Broad adoption
“Architecture of participation”
Hijacked by the community

Great technology

Eclipse is a platform that you can extend through plugins. It is already well-known as a Java IDE, but there are plugins for many more languages (C/C++, Java, Python…). If you remove the language plugins, you are left with RCP, which is a great platform to build rich applications.

Broad adoption

The Eclipse Foundation now has 150 members. Eclipse has about 60% market share in the Java IDE market. Eclipse 3.2 was downloaded 1 million times in the 47 days after its release. The eclipse.org website has 36000 unique visitors / day.

Architecture of participation

The aim of an architecture of participation is to create a “free market for ideas”, by lowering the barrier to entry for contributions. It has been enabled by the plugin system of Eclipse. “You need a cathedral to enable the bazaar!”: the core must be developped in a very rigourous manner, and then there must be a lot a freedom for people developing add-ons.

This is similar to the way Apache httpd is extended by different modules (mod_*) or to Mozilla Firefox extensions, etc. OpenOffice.org is also moving in this direction, thanks to the efforts of people like Laurent Godard.

“Hijacked by the community”

“Hijacked” refers to a book called “Brand Highjack” by Alex Wipperfurth (cool, a new book for me to read ;) ), which advocates “allowing customers to shape brand meaning and drive a brand’s evolution”. Mike, like me, is a reader of the “creating passionate users” blog by Cathy Sierra. He (or she?) says: “Fire the marketers. We are all marketers now”. Marketing is not done by other persons, but by ourselves as technologists.

Conclusion

In conclusion, Mike stressed that Eclipse RCP is the part of Eclipse with the greatest potential and highlighted a few success stories.

Overall, this was a great and inspiring keynote. The Eclipse architecture and process are both a tremendous inspiration for us with our Nuxeo 5 project, where we also have developped a plugin archtecture (using the same underlying technology as Eclipse, OSGi), we aim for clean API for external developpers, we want to help build a community of third-party contributors and extenders though, and of course we are using Eclipse as a Java IDE and as the foundation for our Nuxeo RCP project.

What a month! (Quick reflections two weeks after our Java switch)

We announced our switch to Java two weeks ago. There was of course a lot of preparation before the announcement (besides just coding). That included: creating the new Nuxeo.org website, writing a FAQ, preparing slides, writing announcements for various media (InfoQ, TSS,…), and anticipating some discontent in the Zope community.

Well, two week after going public with the announcement, I’m glad to say that we are completely pleased with the way things went.

We already have released one of the key components of the platform, Nuxeo Runtime and are preparing to release Nuxeo Core this week. Overall, work on the project is going on at a steady pace and we are on track to meet the next milestones of our roadmap.

Our announcement got noticed both by the main Java sites, and countless blogs (including some in languages that we don’t understand without Google Translator).

There has been a little discontent, as expected, in the Zope community, but we mostly got some nice messages of people saying they understood our choice. The most heated discussions took place after Jean-Marc’s post (and its followups) supporting our change.

And, most importantly, we got many subscriptions to the mailing list, which shows there is a real interest in the developer community.

On the business side, we’ve had many calls from system integrators.

I’m especially glad to notice that in two weeks, we got contacts from several new systems integrators, telling us that, now we are using Java, they will be very happy to work with us on new projects. Consequence: we now have a commercial contact with 8 of the 10 leaders of ECM integration in France (some of them the french subsidies of international groups), as well as interesting partnership projects with a big name (I mean BIIIIG) software vendor (more on this another day).

Meet me at Eclipse Summit in Esslingen (or Eclipse Now You Can conference in Paris)

I will be in Esslingen next week for the eclipse Summit.

I’m going to talk about Nuxeo RCP, the rich client side of our Nuxeo ECM project, and about the Apog�e project, which is now an official Eclipse Foundation technology project.

Then I will be at the Eclipse Now You can one day conference in Paris. Similar program for us, except I won’t be delivering the talk myself (you have to share the fun of giving talks with your colleagues, right Laurent ;) ).

I’m looking forward to talk to people interested in our project, so we can discuss how to best collaborate together.

Watch this space for a report on both conferences, and for the slides that are not done yet.

CPS4/Z3ECM sprint, day 3, short report

Yesterday was the third day of the Z3ECM / CPS4 sprint.

Florent and Dario are now able to create complex Zope3 schemas from the JackRabbit schemas. There is now a nuxeo.capsule package that contains just base classes and a nuxeo.jcr package that contains the parser. The next steps are to create new node types in the JCR, export them at startup, and have Zope to parse them as interfaces and schemas. Then we have to find out how to exactly use them.

Jean-Marc took the upcoming Zope 2.10 branch (w/ Philipp's and Jim's branches) and get all the ZCML files parsed and the utilities registered, etc. Then he started to look at Jim's refactoring of local utilities to use the same API. He is now half-way in getting something that works. The next step is to register these utilities in Zope 2.

Michael got the Zope 3 publisher to work in Zope 2 for ZServer.You can now use all the 4 combinations of Zope 2 publisher vs. Zope 3 publisher, and ZServer vs. Twisted.

Tarek has worked on a CPSResourceLibrary, similar to zc.resourcelibrary. It's now finished. Joachim has used it on some portlets. Tomorow Tarek is going to look into the use cases of the JS library packaged on monday in CPS 3.4.

CPS4/Z3ECM sprint, day 2, minimal report

Yesterday was the second day of the Z3ECM / CPS4 sprint.

Michael and Lennart have got Zope 2 working on a Twisted server. They are using the Zope 3 publisher now in Zope 2. No benchmarking has been done yet. The code lives in Sidnei's publication-refactor branch of Zope2 (Sidnei did most of the work).

Tarek, Joachim and Jean-Marc have worked on CPSSkins integration. There were many branches. They finally landed on a Philikon branch where Jean-Marc did all the hard work. Tarek is now working on a "resource library" for Zope 2. Joachim is trying to use the JS library package that was created yesterday manually.

Florent and Dario have written an importer for JCR schemas expressed in CND notation. They have a parser for that and are able to build a Zope 3 schema from it. They are now stuck on the fact that the JCR can have children with the same name.

CPS4/Z3ECM sprint, day 1, short report

The sprint has started yesterday with notable guests Jean-Marc and Dario from Chalmers, Joachim from AixtraWare, and Michael from OpenApp, and a bunch of Nuxeo guys.

The goal is still to provide Zope3 ECM components for CPS4

Yesterday morning was spent discussing the AJAX and JavaScript MVC features of CPSSkins v3 (more info in Jean-Marc's numerous messages in the mailing list, or on the z3lab website).

Then we had a discussion about the JCR content model and the JCR schemas for documents (JackRabbit's CND).

In the afternoon, work has proceeded in pairs or small groups.

Florent and Dario have investigated existing schemas sytems and are trying to find a common subset. Now, they will code and write an importer for the JCR schemas.

Lennart and Michael are looking on how to plug the Zope3, Twisted-based, publisher into Zope2.

Jean-Marc, Tarek and Joachim are working on AJAX and CPSSkins. They have created small package of all the ressources used by cpsskins so that they can be reused in CPS. The goal now is to instanciate de Zope 3 CPSSkins utilities in Zope 2, look for breakages, and fix them.

Bogdan is working on Apogee integration through ICE.

CPS4 project officially started!

Two signs that the CPS4 project has officially started:

Florent has committed a roadmap in the SVN. (This roadmap will obviously be the basis for next week sprint at Nuxeo's office in Paris).
The CPS4 mailing list is now open for the community (and last week's private discussion archives are available too).

Now, you may ask, "what is CPS4?"

In a couple of sentences, that will be expanded in upcoming blog postings: CPS4 is the evolution of CPS3 towards more openness and more scalability. We keep all the ECM functions we like in CPS framework (the repository, the documents schemas, the directories, events, skins, portlets, relations...) , and we leverage both the work done on the Z3ECM project, and the offering of the Apache project (Lucene, Jackrabbit) and a few other established infrastructure Java projects. Last thing: we plan to use the ICE high-performance communication engine for communication between the different components.

FAQ: "Is the CPS 3.4 branch still supported?" R: of course, we also plan to do a CPS 3.4.1 release very soon.

Tag your Trac too!

As you already know, I have started two weeks ago tagging (and asking others to tag) their CPS sites on del.icio.us with the tag "cps-site". I think we have spotted about 80 public CPS sites so far, but I'm sure many are still missing.

BTW: how do you know a website is a CPS site? There are several clues: 1. Look at the icon, if present. 2. Look at the urls (if it contains "section" it's probably a CPS site). 3. Look at the HTTP "Server" header.

Now I have a similar idea: there are many Trac site out there, and Trac is one of the "Python killer-apps". So let's tag also Trac sites with the trac-site tag and see what happens.

Packaging CPS

Packaging CPS is important for ease of installation. Unfortunately, it is also a great deal of effort, but an effort that can be distributed between the original software developers and third-party packagers.

The situation so far for CPS is that we have:

the "big tarball", that installs itself on a fresh Zope 2.9 instance, made by the CPS dev. team.
the Windows installer, which installs Python + Zope + CPS + additional librairies + additional software (converters), made by Ruslan.
the Debian/Ubuntu packages, made by the Debian Zope Packaging Team, and specially Fabio Tranchitella
Gerardo Puerta has made an ebuild for Gentoo.

There are some areas were would benefit from some external help, though:

We are still working on the Mac OS X installer, for which some preliminary code has been written, but I have been told that the current approach is a dead-end and we need to start afresh with a new approach.
There are still not satisfactory RPM packages for common distributions like Fedora, Mandriva or OpenSuse.

Packaging Zope and Zope applications under Linux is challenging, because you have to accomodate both people who just want "one click install" (using apt-get / yum / urpmi), but also provide the basis for rock-solid, possibly multi-instances, Zope server management (including easy instance creation / deletion, log rotation, service initialisations, etc.) without compromising security. The Debian Zope team seems to have done a lot of work to streamline this process for Zope.

Regarding CPS, any help in improving the existing packages, or in making new packages kinds available, would be really appreciated. If you feel in a packaging mood, join the cps-devel mailing list and let's start working together.

Nuxeo recrute

Je viens de poster deux annonces sur Lolix: une pour un (ou plusieurs) d�veloppeurs Zope/CPS (profil: IUT ou ing�nieur, un go�t prononc� pour la d�veloppement d'applications open-source) et une pour un sysadmin.

A CPS project awarded by french magazine "Le Monde Informatique"

Our friend at CEA (French Atomic Energy Commission) have been awarded "best project of the year" in the "public administration" category of the competition organised by the french magazine "Le Monde Informatique" (an IDG publication). They have been using CPS for 3 years and now have 129 instances in production.

According to the article (p. 19), the use of CPS has allowed for 1.9 to 3.8 millions Euros savings over 3 years.

(Note: last year, CPS won 2 awards in the same competition, one for CNCC.fr and a second one for Ardeche.fr! Check for yourself in last year's article).