Table of Contents
This chapter is based on the Debian Subproject HOWTO, which was
written by Ben Armstrong
In this section, issues to think about before starting a Debian Pure Blend will be discussed. It is important to have a clear idea where to head and how to get there before launching into this adventure.
The existing Debian Pure Blends have clearly shown that they depend on a person who keeps things running. If anybody wants to start a project at first, he has to answer the question: "Am I the right person for the job?" Surely this is a question that may be faced with some amount of uncertainty. The way Debian Pure Blends started in the past was for the person with the idea for the project to just start doing the work. After some time using this approach, it became clear that if the project lacked a person to take leadership, the project would become stale. So the initiator has to answer the question clearly, whether or not he is able to continue in the job of leader, considering the amount of time he will have to spend, and the technical and social skills which are needed.
It is as important to decide what your group is not going to do as it is what it is going to do. A clear borderline is essential for the development of the project. Without it, outsiders might either expect more from the project than it can accomplish, or may ignore the project, finding it not helpful because they are not able to find out the purpose.
By maintaining a good relationship with other Free Software projects, some common tasks can be done very effectively. When efforts can be shared, the amount of work for each project can be reduced.
Checking for cooperation with other Debian Pure Blends is always a good idea. In technical terms, this is obvious, but sometimes there are possibilities to share efforts when the goals of two projects have parts in common.
The one who decides to start a Debian Pure Blend takes on a responsibility for this project. It has to be for the good of Debian as a whole, and should bring an extra reputation to our common goal to build the best operating system.
By the time you have begun to think about forming the subproject, have made the commitment to lead it, and have sketched out a bit of where you want to go and how you'll get there, you have likely already done some informal discussion with your peers. It is time, if you haven't already, to take these ideas to the broader Debian developer community, opening discussion on the creation of your group.
At this stage, you will want to reach as broad an audience as possible.
You have carefully thought out what you're going to do, and are able
to articulate it to Debian as a whole. Let everyone know through
list, setting the Reply-to:
<email@example.com> and listen to what
everyone has to say about your idea. You may learn some valuable
things about pitfalls that may lie ahead for your group. Maybe even
show-stoppers at that. You may also find a number of like-minded
individuals who are willing to join your group and help get it
It's all too easy to get lost in ever-branching-out sub-threads at this point. Many people will be firing off ideas left, right and centre about what your group should do. Don't worry too much about containing the discussion and keeping it on track with your main idea. You would rather not squelch enthusiasm at this point. But do try to steer the discussion a bit, focusing on the ideas that are central to your subproject and not getting lost in the details.
At some point, you'll decide you've heard enough, and you're ready to get down to the business of starting your group.
It is fairly important to enable some means for communication for the project. The most natural way to do this is with a mailing list.
Creating a new mailing list starts with a wishlist bug against lists.debian.org. The format of this bug has to follow certain rules.
Before your list can be created, the listmasters will want assurance
that creation of the list is, in fact, necessary. So for this
reason, don't wait for your list to be created. Start discussing
your new project on
immediately. To help distinguish your project's posts from the
large amount of traffic on this list, tag them in the Subject field
with an agreed-upon tag. An example bug report to create the
relevant list is bug
When sufficient discussion on the developer's list has taken place and it is time to move it to a subproject list, add to your wishlist bug report some URLs pointing to these discussions in the archives as justification for creation of your list.
A simple possibility, and one which is fairly attractive because it facilitates collaborative web site creation and maintenance, is to put a page on the Wiki. There is a special Wiki page for Debian Pure Blends.
A good place to put static web pages is the common place for all Blends: http://blends.debian.org. There is a subdirectory for each Blend and it is very easy to create a simple index page there which points to the automatically generated web pages which are mentioned in Section 6.2.4, “Web interfaces”. Following this strategy is quite cheap and has a big effect when using the tools provided by the Debian Pure Blends effort.
Sooner or later a Debian Pure Blend will establish an own project at alioth.debian.org, since this enables many features for group maintaining packages, create mailing lists for different purposes, maintain a version control system like SVN or Git etc. The Alioth project enables to provide web sites as well and the Debian Med project is using this since a long time.
Finally, the best way is to have a page under http://www.debian.org/devel. While not as straightforward as any of the other options, this approach has its advantages. First, the site is mirrored everywhere. Second, the Debian web site translators translate pages into many different languages, reaching new potential audiences for your Debian Pure Blend, and improving communication with other members of your project and interested parties for whom English is not their most comfortable language. Third, a number of templates are available to make your site more integrated with the main web site, and to assist with incorporating some dynamic content into your site. Before you join the Debian Web team you should learn more about building Debian web pages.
Once there is a list, or at least enough preliminary discussion on
debian-devel to get started, and there is some information about the
newly planned Debian Pure Blend available on the web, it is time to
send a formal announcement to
announcement should include references to past discussions, any web
pages and code which might already exist, and summarise in a well thought
out manner what the project is setting out to achieve. Enlisting the
help of fellow developers on irc or in private email to look over
the draft and work out the final wording before it is sent out is
always a good idea.
<firstname.lastname@example.org> have to
be signed by the GPG key of an official Debian developer. However, it
should not be a very hard task if somebody wants to support Debian
while not yet being a developer to find a developer who volunteers
to sign an announcement of a reasonable project. It might be
reasonable to send this announcement also to other relevant non-Debian
lists. If your announcement is well done, it will draw a number
of responses from many outsiders, and will attract people to Debian.
While there are a variety of different kinds of work to be done in Debian, and not all of them follow this pattern, this document describes one particular kind of project. Our discussion about Debian Pure Blends concerns sub-setting Debian. A sub-setting project aims to identify, expand, integrate, enhance, and maintain a collection of packages suitable for a particular purpose by a particular kind of user.
Now, strictly speaking, a subset of packages could be more general than described above. A subset could be a broad category like "audio applications" or "network applications". Or it could be more specific, such as "web browsers" or "text editors". But what a sub-setting project such as Debian Jr. aims to do is not focus on the kind of package, but rather the kind of user. In the case of Debian Jr. it is a young child.
The sort of user the project looks after, and which of the needs the project hopes to address are defined by the project's goals. BA: I had a bit of trouble deciding how to punctuate the following passage. I considered and rejected the advice given here in response to Kirsten's question about punctuating a list of questions: http://www.udel.edu/eli/g20.html, instead following the advice found elsewhere on the Web that double punctuation should be avoided. Thus, Debian Jr. first had to decide which children the project would reach: "What age?" "English speaking children only, or other languages as well?" Then the project had to determine how and where they would be using Debian: "At home?" "In school?" "Playing games?" "On their own systems?" "On their parents' systems?"
The answers to all of these questions are not straightforward. It is very much up to the project to choose some arbitrary limits for the scope of their work. Choose too broad a focus, or one which duplicates work already done elsewhere, and the energy of the project dissipates, making the project ineffective. Choose too narrow a focus and the project ends up being marginal, lacking the critical mass necessary to sustain itself.
A good example was the request to split the microbiology related packages out of Debian Med into a Debian Bio project. This is reasonable in principle, and should really be done. In fact, the initiator of Debian Med would support this idea. So he gave the answer: "Just start the Debian Bio project to take over all related material. Until this happens, Debian Med will cover medical material that deals with sequence analysis and so forth." Unfortunately, there was silence from the Debian Bio proponents after this answer.
Of course, it sometimes turns out that you start working on a project thinking you know what it is about, only to find out later that you really had no idea what it would become until the user base has grown beyond the small community of developers that started it. So, none of the decisions you make about your project's scope at the beginning should be taken as set in stone. On the other hand, it is your project, and if you see it veering off in directions that are contrary to your vision for it, by all means steer it back on course.
According to the plan of the project, the first metapackages (Section 6.1, “Metapackages”) should be developed. It is not always easy to decide what should be included, and which metapackages should be built. The best way to decide on this point is to discuss on the mailing list some well thought out proposals.
Section Section 6.2.2, “Text user interfaces” mentions tasksel as a tool to select a Debian Pure Blend, and explains why it is currently not possible to get a Blend included into the task selection list.
Besides metapackages, you may want to develop some new packages for your blend in order to customize some system behavior, like changing XDG menu style. Those normal packages are not defined in your tasks, hence we need to define it properly so that blends-dev will know how to package these new, normal packages. Adding new normal packages is almost identical to debian packaging, just that you need to define your package parameters in debian/control.stub instead of debian/control. Section Section A.3, “How to develop new normal packages in Pure Blends” describes what you need to know for adding new normal packages in Debian Pure Blends in more detail.
Beyond the release announcement for Debian itself, it is necessary to put some thought and work into a release announcement for the first release of a Debian Pure Blend. This will not only be directed at the Debian developer community, but also at the users. This will include potential new Debian users abroad, who may not be on a Debian mailing list. Here, the same principle applies as for the first announcement of the project: it is important to consider sending the information to other relevant forums.
By this time, people have newly installed Debian along with the material in the Blend, or have installed the metapackages on their existing Debian systems. Now comes the fun part, building relationships with the user community.
Users are a mixed blessing. In the first development phase there are some developers who are users, and some intrepid "early adopters." But once it is released, the first version is "out there," and the project will certainly attract all kinds of users who are not necessarily as technically savvy as your small development user community. Be prepared to spend some time with them. Be patient with them. And be listening carefully for the underlying questions beneath the surface questions. As draining as it can be to deal with users, they are a very key component to keeping your development effort vital.
Should a user list be created? It's not as cut-and-dried as it might at first appear. When user help requests start coming in, you might at first see them as a distraction from the development effort. However, you don't necessarily want to "ghettoize" the user community into a separate list early. That's a recipe for developers to get out of touch very quickly with the users. Tolerate the new user questions on the developer list for a while. Once a user list is finally set up, courteously redirect user questions to the user list. Treat your users as the valuable resource about how your project is working "in the field" that they are.
Fortunately, we're not in the business of supporting users alone. Look beyond Debian for your allies in user support: Linux user groups (LUGs) and the users themselves. Develop an awareness of who has stakes in seeing your project succeed, and enlist their help in getting a strong network of support established for your work.