By Nick Kew
I started working on a range of Web applications in the 1990s, the first of them internal to my (then) workplace where it provided an operator interface to the daily processing, archiving and distribution of satellite image data; the second a forerunner of what is now called social media, and my first use of an Apache server. The release of Apache HTTPD 2.0 drew me from server user to developer: in part because I needed to re-implement some existing functions, but more excitingly because I saw tremendous potential for the server itself to become a powerful applications platform. This led me to working on the core software and interacting with the Apache community alongside releasing my own modules and documentation. In 2003 I gave my first ApacheCon presentation, and sometime after that was invited into the Foundation first as a Committer, and became a Member in 2005. Since then my interests have encompassed not just the Web server and related projects, but also the Apache community and its dynamics. I’ve been involved in mentoring several projects through the Incubator. If you were to ask me today about the single goal I’d most like to accomplish, it’s a framework for Identity management that is not merely cryptographically strong, but sufficiently straightforward for the world to use, and robust against social engineering attacks such as phishing, while at the same time free of any centralised authority (such as government) whose motives might come under suspicion. An end to identity fraud, and to password management nightmares.
Much has been said and written about The Apache Way . It typically focusses on the importance of community, and on the democratic and meritocratic elements of project governance. And on the role of an Apache project’s formal governing body, the Project Management Committee (PMC), comprising contributors elected by the community on the basis of their track record of contributions and constructive engagement.
In practice, the role of the PMC is largely reactive. The big, interesting questions like "what direction does our project go from here" are discussed in public, where inputs from the wider community are welcomed. A lot of the nitty-gritty detail is determined by the needs of the community as measured by feedback received and seen, and by the needs of individual developers. The latter is often described as "scratch your own itch".
Setting aside the purely reactive, the detail of what actually happens is often determined as much by what a developer is prepared to work on than by any grand plan. If you do the work, then you control what happens. If you need something, you go ahead and make it happen. This has occasionally been described by a clumsy English-Greek hybrid word "do-ocracy". In the hope that it’s not too late to nip that in the bud before it becomes as ubiquitous as the Greek/Latin hybrid of which CP Scott memorably told us no good would come , I shall instead call it Pratocracy. Rule of the Makers. It applies both within an Apache core team and in the broader community.
So how does a Pratocratic project work? How does it avoid the chaos of a Tower of Babel? How does it remain cohesive and focussed? And if an Apache project looks like a good basis for your company’s needs, how best can you work with the Pratocracy towards your goals?
Taking the first question first, how does the Pratocracy work? Eric Raymond proposed the contrast of the Cathedral vs the Bazaar: the autocracy of the traditional project vs a free ecosystem. The Bazaar ecosystem is essentially Darwinian, in that the strongest projects prosper and evolve, while a Long Tail go nowhere very much and are forgotten as their authors move on to new things.
The Apache processes are all about ensuring our projects are fit to be among the winners in that ecosystem. An Apache project is a graduate of an incubation process  that requires it to build a broad development community, with sufficient diversity to survive the loss not merely of a key individual, but even of a key company team that might have been the originators of the project. The key principle is that if the community is healthy, the software will follow.
Thus the context of the Pratocracy is selective: it is the well-known Meritocracy element of Apache. Members of the core community are elected based on demonstrated commitment to the project. That means they understand the distinction between Apache work and company work, and are comfortable holding multiple allegiances and avoiding or resolving potential conflicts between them.
So, to the crucial question. Suppose I have identified an Apache project that doesn’t fully meet my company’s needs, but is the ideal starting point for our project. How do I best work with the Pratocracy to make it happen?
Let’s consider some possible approaches:
- Ask via the development list and issues database.
- Approach developers individually to discuss the project.
- Just go ahead with the company’s in-house developers.
Each of these approaches has merit in its own right. But in many cases, the ideal approach will encompass all the above. In an Apache community, the whole is greater than the sum of the parts. The closer you can work with the community, the more benefits you stand to derive, particularly if your project finds itself wanting to integrate more deeply with the Apache project than is provided for in stable public interfaces, when the alternative to working within the Community is a patched core and the risk of a maintenance nightmare. On the other hand, there are good and bad ways to engage with an Apache community: the right approach can make all the difference to the benefits you can derive from it.
First, it’s always worth asking about your needs. Post a question about your requirements to the project’s lists, or issues database. Engage in discussion. It may be that someone somewhere has already done the work, or that it’s closely aligned with someone’s work-in-progress. Even if you draw a blank, it’s a start to engaging with the community.
Approaching individuals is most likely to be appropriate if you’re looking to hire in expertise. General questions are likely to be redirected to public support fora, but some Apache developers work as consultants and may welcome a professional approach. They can help you figure out the best way to proceed, both technically and in terms of working with the Apache community. But bear in mind the potential for conflicts of interest. If you want your work to happen within the Apache project, your consultant will want to be satisfied that you and your project are indeed a good fit, and won’t cause trouble such as unmaintainable software down the line. And of course, any NDA should make it clear you’re not laying claim to the consultant’s regular Apache work!
If you have the in-house developers for it, you might just task them to go ahead with the work. That’s easiest when you can work with public, stable interfaces of the Apache project, harder and messy if you have to hack into the core. When you start that work, the Apache community becomes your team’s best resource, so if your developers are any good (and not cowed into keeping their heads down and never talking to outsiders) they’ll be interacting with the Apache community as they go along.
Now you’re participating in the Pratocracy, albeit (usually) outside the core community. You have something to offer. If you share it with Apache, members of the Apache community will take a look, and may get involved. At best this can lead to a virtuous circle where your developers not merely benefit from the Apache team, but contribute to it, and earn themselves the merit to be invited to join. But this is by no means automatic: both developers and their managers need to be able to see outside corporate control structures. Crucially, individual developers have to be comfortable separating out their Apache hats from their professional allegiances, which means managers have to allow that to happen. But don’t try to manage such a process from above: that’s herding cats. Just facilitate!
That probably means that at some stage you’ll have to sit down with the team and fix boundaries: what belongs in the Apache project vs what belongs in your company’s own space. It helps immensely with developers’ confidence if they’re not forever looking over their shoulders and wrestling with ambiguity. This may be obvious if your project isn’t destined to be shared openly with the world, but it’s a discussion worth having even if everything is Apache-licensed. In the latter case, the boundaries will focus on what your company expects to sell and/or support commercially and wants to maintain full control on, vs what passes out of your direct control.
Even if your company merely clones or mirrors the Apache project and contributes everything, you’ll want to separate out matters of your business model, and customer-facing vs Apache-facing matters.
Note that it’s easy to get this kind of thing wrong. Something we sometimes see at Apache is developers who’ve been told by their managers to join an Apache community, but have no specific tasks and no motivation. That becomes painfully obvious, and leads nowhere useful.
The right approach is to give your developers real-world tasks and let them come to the Apache community – if their course takes them that way – in their own time. If close interaction with the Apache project is an important goal, you may want to approach and maybe hire individual developers having a strong track record with Apache or comparable Open Source environments.
To conclude, let’s consider a real-world example from my own experience with HTTPD (the Web server still commonly called "Apache") and APR projects. The history of SQL database integration into the Web server spans many project teams, both within and outside Apache, both corporate and community-driven. Each developer or team just went ahead with something appropriate to its time, and (at least those you or I have heard of) achieved success in their respective contexts.
In the early days, there was a Web server with no SQL support. So any application that required it had to manage its own database connections. A range of third-party modules were developed for purposes such as authentication by SQL, and logging to SQL, both with popular databases such as MySQL and PostgreSQL and in commercially-supported products with commercial databases. Generalised frameworks for Web/SQL applications were provided by scripting modules, pioneered by mod_perl, giving developers a language’s own database interfaces, together with a persistent database connection.
In time, the server and its applications outgrew this ad-hoc approach, with multi-threaded processing and multiple applications in a server. Building on a range of ideas, a native SQL database connection framework was proposed and developed, offering some major efficiency gains to applications through dynamic connection pooling and sharing. It also promoted rationalisation, so in an area like authentication, a whole raft of different database-driven modules could be replaced by a single module for generic SQL authentication. Different SQL databases are handled by per-database drivers which serve all kinds of application in the server.
This work took place within the core Apache team, and propagated more widely as others such as the scripting languages incorporated it as an option for their users. It evolved from minimalist beginnings to provide more capabilities as required by developers "scratching an itch", and has benefited from contributions of database drivers from a range of different sources.
Having said that, it remains intentionally fairly small, and won’t meet the needs of every application. Some projects continue to use SQL their own way. The fact there is now a Right Way to access SQL doesn’t mean that other ways are wrong! There are still third-party projects that, for their own reasons, use other ways to access SQL. The fact that someone has done the work and met a need gives them an entirely valid niche in a Pratocratic ecosystem.
The bottom line? Many Apache users can just use our software off-the-shelf (perhaps after asking around for solutions to problems), others may need to add significantly to it to meet their needs. If you fall into the latter camp, just go ahead and do it: you can discuss and/or share your work at any point you choose! The basic management decision is the tradeoff between control of your work and community input. Interaction with the project’s core people is down to how your developers work best. If leaving that to chance is not sufficient –perhaps because you would like to influence the project’s core software or introduce new APIs– that’s the point when you’ll need to consider your developers’ track records in community-driven Open Source, and make sure you have that expertise to hand.