Diff for /wikisrc/users/jdf.mdwn between versions 1.6 and 1.10

version 1.6, 2012/07/06 22:47:31 version 1.10, 2013/02/16 12:37:54
Line 1 Line 1
 [[!toc ]]  **Contents**
 # jdf's wiki page  [[!toc levels=2 ]]
 This is just my wiki page where I want to write down my goals, project ideas, running projects and what I generally want to do.  
 # project ideas  
 These project ideas are not sufficient for the [[projects]], perhaps only my own stupid ideas, but what I would like to see for NetBSD. Some of them might be redundant with the official ones, but these are the ones I might personally take care of one day.  
 * zfs -- there was an attempt to port zfs to NetBSD, but it stalled. Somebody to take care of zfs being imported to NetBSD, and running smooth, would be *very* nice.  
 * HAMMER -- I don't know much about it (see DragonFly BSD), but from what you can read, it seems very nice. Looking at the current state of Oracle and zfs, it might be better to care for HAMMER than zfs. On the other hand, zfs is very stable and widely distributed.  
 * unionfs -- there are some bugs with union I would like to see fixed. Imagine running a live CD by having a root read-only, and then simply mounting a tmpfs writable upon that.  
 * raidfs -- once unionfs is fixed, it could be possible to integrate a mirror, perhaps some redundancy checks with unionfs to have a flexible raid on vnode basis, not depending on the underlying filesystem or device.  
 * userwiki -- I know this is a controversary topic... So no matter how and where, but having a place where users can contribute content in an ordered way (i.e., not on a mailing list), would be nice. See below for advocacy.  
 * NetBSD derivates/distributions -- provide prepackaged NetBSD distributions for several applications, see below  
 Providing a more unified, complete system:  
 * binary updates -- imho this is being worked on, but I don't remember who it was. Like the freebsd-update tool from FreeBSD, an easy way to update NetBSD base system from binary. Currently, you have to unpack manually and run [[!template id=man section="8" name="etcupdate"]], or use [[!template id=man name="sysinst" section="8"]] for this purpose. Running something that fetches the necessary updates and updates these files would be nice.  
 * disk utility -- a tool for unified disk editing, and not having to use [[!template id=man name="disklabel" section="8"]], [[!template id=man name="gpt" section="8"]], [[!template id=man name="fdisk" section="8"]], and [[!template id=man name="dkctl" section="8"]] separately, which can be very confusing, would be very nice. This shouldn't be too difficult. The functionality and the code is there, you just had to think of a usage, rewrite the frontend, and provide a compatibility mode for the old tools. Perhaps gpart is what I'm looking for, but I didn't dive into it.  
 * automatic NetBSD installation -- important for larger setups, where you want to automatically install NetBSD. Essentially, it should just be a small script setting up a ramdisk, downloading tarballs from a configurable source, partition disks (no hassle - just take the whole disk), make filesystems, extract tars, run installboot, execute manual scripts to be downloaded somewhere (e.g. if the user wants to setup network configuration or root password)  
 * a tool to change values in rc.conf, and enable/disable services. Currently, there's a bunch of scripts to check, then edit rc.conf, etc. Tom Rhodes from FreeBSD had ideas towards this (together with fscd), to build a more unified rc on top of current BSD's rc, just as an add-on. I'd really like to see that real, and for NetBSD, some time.  
 I won't be working on this, as I'm not really into such low-level things:  
 * run NetBSD in a web browser. There was a project which ran Linux in a web browser. If it's possible, NetBSD should run in a web browser, too.  # jdf's wiki page
 # My personal interests  
 I'm not into pkgsrc, but for the easy packages I use regularly, I try to help updating them.  
 * fixing small userland bugs  
 * caring for [[!template id=pkg categ0ry="sysutils" name="fscd"]] (a daemon to check service run state)  
 * writing mkdumpdisk - a tool to backup the system, including restore  
 * caring for [[!template id=pkg category="www" name="opera"]] being up-to-date  
 * caring for [[!template id=pkg category="devel" name="fossil"]] being up-to-date  
 * keeping the events site up-to-date  
 # NetBSD derivates  
 Currently, NetBSD is a very generic operating system, leaving almost all choices up to the user. While some consider this a strength, and it definetly is for people who know what they're doing, it's an obstacle for people who then have to setup *everything* by hand.  
 It shouldn't be much work to just package distribution sets that already include a list of packages it installs and several preconfigured configuration files, maybe also some additional wrapper scripts.  
 Technically, this could also be integrated to pkgsrc as well. You just have to package everything you want (dependencies, files, etc.) into one package, and then depend on this single package. The finalized derivate would then just depend on this single package, and maybe contain all needed files already on the CD itself, as well as some configuration files for /etc.  
 * NetBSD serve -- a compilation of tools and scripts that are useful for using NetBSD as a server. Maybe there could be even more flavours like this, packaged into single pkgsrc packages, which contain preconfigured "appliances" like an ldap package, a webserver package, etc., everything with a sane configuration default.  
 * NetBSD develop -- just a compilation of some desktop tools. Telling somebody what he has to do to get a graphical environment with NetBSD (light-desktop) is annoying, so just have a distribution where light-desktop and maybe some other tools are preinstalled. I think 'develop' might be a good description of the targeted audience, as then you can choose a fairly tight set of packages to put in: light-desktop, vim, several VCSs, etc.  
 * NetBSD secure -- there was this teenie book, "Little Brother" from Cory Doctorow. There, they had an operating system that was actually only on a CD, and encrypted the whole hard disk, and everything ran on the hard disk. Imagine you had the same for NetBSD... A base system (Jibbed would be sufficient), which has some small wrapper scripts trying to decrypt a hard disk, and if it succeeds, mounts this as its /usr/pkg or so. You could have a completely virgin OS (the CD, which is read-only), and a completely encrypted hard disk, which carries everything variable. Updates of the base system would be just burning a new CD... I know there are close projects already done in Linux, but they all assume an USB flash drive.  
   Note: This is not what I'm really working on, it's just a place to gather some 
   notes I took about some topics.
 # Advocacy  ## NetBSD flavoured
 Imho, NetBSD currently doesn't have a good visibility in the Open Source community. These are my observations (if I knew solutions, I would write or apply them ;-):  Currently, NetBSD is a very generic operating system, leaving almost all
   choices up to the user. While some consider this a strength, and it
   definetly is for people who know what they're doing, it's an obstacle for
   people who then have to setup *everything* by hand.
   Creating a *NetBSD flavoured* distribution shouldn't be much work, and require 
   just minor sysinst modifications.
   It shouldn't be much work to just package distribution sets that already
   include a list of packages it installs and several preconfigured configuration
   files, maybe also some additional wrapper scripts.
   On the other hand, you could also add some package calls to sysinst and just 
   provide a list of packages you consider necessary.
   My original attempt was to create a range of distributions for different 
   purposes, i.e. one for developers, one for graphic designers, one for servers, 
   etc. I don't know if this is the right way, esp. since some of the applications 
   are *very* specific. You cannot really provide a sane server default 
   installation except for some basic things like installing a vim, but that's all.
   My current idea is to provide just one, maybe named *NetBSD flavoured*, with at 
   least the following tools on board:
    * vim
    * pkgin
    * git
    * fossil
    * subversion
    * some other important VCSes
    * light-desktop (i.e., LXDE)
    * screen (tmux is in base)
    * some sane X terminal emulators
    * a browser (Firefox?!)
    * a mailer (Thunderbird? Claws-mail?)
    * emacs (maybe too large?)
    * perl
    * python
    * mplayer (when it's possible to pack it up)
    * pdf viewer
    * preconfigured bozohttpd running on localhost showing documentation
   ## NetBSD documentation
   In [this 
   I shared some ideas about what to do with documentation. Though much of it 
   was proven not practical by the replies, I still have one idea: Unify 
   documentation of NetBSD, and provide it all on a NetBSD system.
   The first step is to merge as much content as possible into the NetBSD wiki. 
   Currently, the NetBSD documentation is very diverse in its distribution form.
   Then, the Google Code-In produced some nice results, including a CGI for a small 
   markdown wiki to browse the wiki (if it was offline), and maybe even a terminal 
   markdown browser.
   Finally, ship these two in a pkgsrc package or even with base, and provide a 
   small script which regularly updates the documentation.
   ## NetBSD website
   Currently, the NetBSD website is written in HTML and Docbook and requires many 
   tools to be edited and committed. The final goal should be to have just a small 
   homepage with a bit important information, but all the essential technical 
   information should be in the wiki. There's also a separate page for this: 
   Though the plan is currently to migrate *all* contents to the wiki, I don't 
   think this is the way to go. A wiki just doesn't leave a good impression.
   ## NetBSD community and marketing
   Just some thoughts... I think NetBSD has a very bad way of making technical 
   ecisions which are counterproductive from a marketing point of view, or just are 
   not used for marketing purposes.
   The world has changed; nowadays, there's a growing *hacker community* which 
   consists of many people with an age below 30. They're just not used to the 
   flexibility of the old tools Unix provides, and to the flexibility you have 
   with a modern Linux.
   There are repeating questions why NetBSD doesn't use git as its primary VCS, but 
   rather CVS. CVS *is* indeed a very mighty tool, but many people don't know. They 
   like git more because they can explicitly `push` with it (and don't know about 
   hooks in CVS or Subversion).
   The same holds for many other decisions.
   NetBSD has a very... oldish view of how a community should be organised. On the 
   one hand, there are the developers, which are coding the project, maintaining 
   the website, maintaining packages, maintaining documentation, organising events, 
   organising NetBSD itself... and on the other hand, there are the users. They're 
   rather consumers than contributors.
   The few ones which want to contribute are doing so, and after some time becoming 
   developers with the right and possibility to do everything, but there's nothing
   in between. There's only few community involvement overall, though there are
   many topics which don't require a developer status.
   I think breaking with the old habits and providing more community involvement 
   and community support is the way to go, but except for starting with a 
   user-editable wiki, I don't have many ideas how to do so.
   ## NetBSD current
   The same problem exists imho with the release cycle. The standard release cycle 
   of NetBSD is too slow for many people who use it privately (just see how 
   wide-distributed Arch Linux got), and tracking current is a rather obscure thing 
   with compiling things on your own, etc. ...
   And it's not well-documented. There *are* changes, but who knows them? Which was 
   the current version where tmux was imported? Etc.
 * I have seen several smaller software projects which provide compiled binaries on their website, in many cases there was just a reference to "look into your system's package management, it's already there." For nearly all of them, the package is in pkgsrc, but the website doesn't reference it.  Tracking these changes more centrally, and providing a nice way to install and 
 * Though NetBSD is represented at most of the larger Open Source events, the advocacy material could be updated. In central Europe, there are already new flyers, badges as giveaways, T-Shirts and badges to sell, but there could be more. If you have ideas, write to netbsd-advocacy (at) NetBSD (dot) org.  track a current installation would be a great benefit for NetBSD.
 * There are many nice projects within NetBSD, but they aren't very visibile from the outside. [[!template id=man name="rump" section="3"]]? [[!template id=man name="npf" section="3"]]? The new [[!template id=man name="apropos" section="1"]]? All the prestigous projects are often kept silent, commited, used, but most people outside the NetBSD community (and especially *BSD) know about them.  
 * For contributors with "minor" interests like writing small articles, correcting manpages, and providing other types of small patches, it seems extraordinarily difficult to get things done. A user-commitable wiki, and some methods to improve developer-contributor communication, might help here.  

Removed from v.1.6  
changed lines
  Added in v.1.10

CVSweb for NetBSD wikisrc <wikimaster@NetBSD.org> software: FreeBSD-CVSweb