(no commit message)

[wiki.git] / projects / ideas.mdwn

[[!meta title="Project Ideas"]]

A lot of what SIPB does which is not helping users with immediate
problems and answering questions falls under the heading of SIPB
projects. SIPB projects are usually things that a member or two have
decided are worth doing and are working on. The main qualification of
SIPB projects is that they help the MIT community or the world at large
in some way. In general they also involve computers.

Here's a list of SIPB project ideas that have been mentioned recently,
along with the person who's suggested the project or a team that would
be good to contact. Feel free to get in touch if something sounds
interesting or you want advice getting started.

A [[list from 2008-2009|/doc/project-ideas]] may also have some relevant ideas.

[[!toc]]

## "add me to this list" button

If I'm a webmaster for some group with an announcement list, it would be
really nifty if I could add some HTML code to my page to create a button
that, if someone clicks it, would use MIT certificates to find their
username and automatically add them to a list. This isn't particularly
difficult; you'll need to get a keytab (essentially, an account in
Kerberos / Moira) for the website, and write a script to set that keytab
as the list's "membership ACL" and a webapp to use that capability to
add or remove people from the list.

[adehnert notes 2013-02-24: More recently, WebMoira has been redone, and might more conveniently support this. It might be possible to directly integrate with WebMoira, or you might conclude that WebMoira is already good enough.]

_Contact: geofft_


## debdiffs of Debathena packages

We have [a webpage](http://debathena.mit.edu/package-list/proposed) to
list all Debathena packages in the ["proposed"
repository](http://debathena.mit.edu/experimental#proposed), i.e., things
we just changed and are waiting a few days for testing before pushing to
the clusters, etc. However, that page gives you no information on what
the change was. A very useful addition would be to use the `debdiff`
tool to link to the changelog and code changes for each new package.

_Contact: broder, debathena_

## Better Trac-email integration

A bunch of SIPB projects use [Trac](http://trac.edgewall.org/) as a bug
tracker. The current email adapter we use [isn't very
clever](http://debathena.mit.edu/trac/ticket/308); it'll send
out each update to a ticket with a generic subject line, so it doesn't
easily indicate whether the bug was resolved or someone just commented
on it. It also doesn't know how to receive e-mail, so we can't reply to
the e-mails it generates and have our comments go back into the bug
tracker, and we can't [make bugs@mit.edu create Trac
tickets](http://debathena.mit.edu/trac/ticket/216). There are one or two
alleged plugins to do this, but they create a new ticket on every
e-mail, rather than doing something intelligent with replies; a better
plugin in both directions would be extremely helpful.

_Contact: debathena, geofft, broder_

## PulseAudio support for the office sound system

There's a server in the SIPB office that you can "print" MP3 files to,
and they will get played one by one. A significantly more modern
interface to this would be to make a
[PulseAudio](http://www.pulseaudio.org/wiki/AboutPulseAudio) server out of the
computer hooked up to our sound system; this lets computers running
PulseAudio (notably, most recent desktop Linux distributions, but
Windows computers that install Pulse as well) route their sound to the
office speakers.

_Contact: geofft_

## Cluster and printer viewer for the iPhone

You might have seen the xcluster displays (in the SIPB office, near the
four building 11 quickstations, and in the building 12 cluster) that
show you which clusters aren't in use and which printers aren't hosed.
This would be immensely more useful as an iPhone application.

_Contact: ccpost_

## Snapshotted virtual machines for all popular Linux distributions

A common need for people writing Linux software is to check that their
software works on many popular distributions (Red Hat, SuSE, Debian,
Ubuntu, Xandros, etc. etc.), and on multiple versions of these
distributions. The qemu system emulator, and its virtualization cousin
kvm, support image snapshots and a fairly lightweight process for
running virtual machines, so you could set up an Athena locker with a
few commands that give you a temporary copy of a Linux system for a
particular distro and version that runs inside your terminal.

_Contact: geofft_

## Efficient disk snapshot support

The previous project and a lot of other work could be made significantly
more efficient by rethinking the design decisions in qemu's snapshot
support. If this is the kind of thing that interests you...

_Contact: nelhage_

## SIPB PostgreSQL server

A lot of people prefer Postgres to MySQL, so having a community Postgres
server akin to [sql.mit.edu](http://sql.mit.edu/), with similar access
control, could be a useful project. You could set it up on
[XVM](http://xvm.mit.edu/) with an eye to redoing the configuration on a
scripts-hosted VM at some future point.

This could easily become a bigger project, integrated with scripts'
autoinstallers, etc., depending on ambition. Or it can just be a
standalone server for people who want access to a Postgres database, but
don't want to set up and maintain a server on their own.

_Contact: mitchb, geofft_

## Rewriting "add", "attach", etc. for Debathena

Athena used to have a system _liblocker_ for managing the /mit
directories and attaching and detaching links to networked filesystems.
Debathena and Macathena come with an automounter for /mit, so links
automatically appear when you use them, and nobdy cares much about
detaching them, but we still use most of the complexity of liblocker
because we haven't gotten around to cleaning it up. There's a [design
proposal](http://debathena.mit.edu/trac/wiki/FixingLiblocker) on the
Debathena bug tracker listing what should be a better implementation.

_Contact: broder, debathena_

## [Checking scripts.mit.edu servers for consistency](http://scripts.mit.edu/trac/ticket/84)

Now that we have five or six web servers (I've lost count), it's
become entirely too easy to change something on one or some but not
all of the servers. We often test changes, like newer versions of
packages or tweaked configuration, in place on one of the servers, but
we'd like something to remind us if we don't copy these changes to all
the other servers. This involves checking packages (RPM, perl, etc.) as
well as config files in /etc and possibly other things like LDAP config.

I've also been told by multiple people that
[Puppet](http://reductivelabs.com/products/puppet) or some other
configuration mangement framework is the Right Answer here. We looked at
Puppet and a bunch of others last summer and concluded none fit our
workflow well, but we could re-evaluate that.

*Note:* mitchb has begun working on this. Coordinate with him if you'd like to
work on this project.

_Contact: geofft, scripts-team_

## SIPB Library

SIPB has a bunch of books in its library. It'd be nice if a list of
the library books also existed online in some sort of sane, searchable
database. One possible platform is the [Exhibit](http://simile-widgets.org/exhibit/) project (which originates from a collaboration between the Haystack group in CSAIL and the MIT Libraries). This would require mostly just making a spreadsheet of the information. Check out <http://sipb.mit.edu/library/> for the current state of the catalog.

_Contact: zhangc, fawkes_

## Improve the Setup and UI for new users of Zephyr

Currently, it is a pain to get someone else set up using zephyr
within screen on Linerva with automatically-renewing tickets.
We should write scripts to set them up so that that all they
have to do is enter a command or click an icon, type their kerberos
password, and then know how to use Barnowl.

_Contact: afarrell_

## Build a web client providing full control to the user

Create a web browser extension (or possibly a stand-alone browser) that is "web-developer antagonistic".  It essentially takes full advantage of the
immense power of a web client, ignoring the wishes of the web developer and server while letting the user take full control.  Every JavaScript event (ideally variable, function, etc.), every rendering decision, every cookie setting, every HTTP request, if the user so desires, etc. is not only made visible but easily changeable.  When making a request, every aspect of the
transaction can be completely fiddled with.

There are several extensions that allow you to do stuff sort of like
this (Firebug, Web Developer Toolbar, Chrome's web developer tools).  But they're hard to use and really mostly just debuggers.  Our goal here is to build something that goes out of its way to give you complete control via a nice UI.

_Contact: leonidg_

## Improve git with shared checkouts

Around SIPB we're kind of [big](http://sipb.mit.edu/iap/git/) [fans](http://web.mit.edu/cluedumps/slides/understanding-git-2008.pdf) [of](http://blog.nelhage.com/2010/01/on-git-and-usability/) [git](http://negativespace.mit.edu/2010/03/08/gitionary-the-graphical-game-of-git-guessing/). But there is an area that git comes up short. We have a lot of common directories where people really just want to edit files in place (instead of wanting to clone/checkout, edit, commit, push...), but git doesn't support that well. It would be cool if there was a way to work with non-bare repositories in shared directories.

One idea might be using FUSE to present a separate checkout to each person using the directory.

_Contact: broder_

## Scripts Pony Improvements

[Scripts Pony](http://pony.scripts.mit.edu) is scripts.mit.edu's new hostname management system.  It was just released recently, and has lots of bite-sized improvements remaining to be implemented.  Particularly good ideas include adding the ability to show and edit hostname aliases, checking whether hostname paths exist and giving appropriate feedback, creating a zephyrbot to allow people to approve tickets easily, and adding the ability to check hostnames in Moira automatically.

See [the project TODO file](http://web.mit.edu/pony/TODO) for more ideas.

_Contact: xavid_

## A zephyr log viewer

Many SIPB-affiliated people use the [Zephyr](http://zephyr.1ts.org/) messaging system, and the [Barnowl](http://barnowl.mit.edu/) client for it in particular. Barnowl has a number of very nice features that make it easy to follow large amounts of zephyr traffic: search, color coding, auto-narrowing, etc. Barnowl can also store logs of zephyrs sent and received for future reference, but the logs are saved separated by class in a way that's quite annoying to navigate sometimes. A Barnowl-like interface (perhaps implemented as a Barnowl plugin) for viewing zephyr logs would be a great thing to have.

_Contact: oremanj_

## Multiple views in BarnOwl

[BarnOwl](http://barnowl.mit.edu/) nominally has views support. However, it consists of verifying that the view name is "main" and returning an error otherwise. It would be nice to maintain multiple sets of view state at once. This would involve figuring out semantics, moving some data structures around, adding the new functionality, designing some interface, and probably much testing for subtle bugs.

_Contact: davidben_

## Etherpad

[Etherpad](http://etherpad.com/) is an awesome tool for online collaborative text editing. It's recently been open-sourced; set up a Java servlet container on XVM, make it work, and then start adding cool MIT features like, oh, the ability to edit daemon.etherpad-writable files in AFS, login with certs and see all your files, and print to Athena printers. Or add cool non-MIT features such as an Emacs client (possibly proxying [Infinote](http://gobby.0x539.de/trac/wiki/Infinote/Protocol), which appears to have some F/OSS implementations already), or integration with [codepad](http://codepad.org/) or [gists](http://gist.github.com/).

*Note:* tvald has set up [etherpad.mit.edu](http://etherpad.mit.edu/).
Coordinate with him if you'd like to get features into etherpad.mit.edu.

_Contact: geofft_

## RFC pretty-printing

I think every RFC website I know of just puts the entire RFC plain text in a &lt;pre&gt; tag. It should be possible to parse out the headers and footers every "page", make section headings into useful ones, and reflow the non-diagram text in a proportional font, which would make RFCs more readable.

_Contact: geofft_

## scripts.mit.edu file upload form

Really, it should not be as hard as it is to make a website where people can upload files too big for e-mail to your locker. Either find a good, simple, PHP or Python or something file upload app, or write a (relatively small) one of your own and a scripts autoinstaller.

_Contact: geofft_

## fakechrooted distros in AFS

Thanks to the deeply disturbing magic of a couple of programs named fakeroot and fakechroot, you can without root privileges enter a "chroot" of an operating system and look at how things work in it. Create some appropriate chroots for popular Linux distros in a locker and appropriate wrapper scripts.

_Contact: geofft, broder_

## A couple of C/C++ hacking projects

* Each AFS cell has its own database of users and groups. If you run `ls`, it will look up users and groups against the local machine's conception of users and groups, so if you take a stock Linux etc. machine and look at most any AFS cell, you'll get a bunch of unhelpful numbers. Make an interface that stands a decent chance of being merged into upstream `ls` to permit it to call `pts examine` (or, rather, the AFS library equivalent) against the appropriate servers instead of `getpwnam` etc. on AFS files. See also [Debathena Trac #300](http://debathena.mit.edu/trac/ticket/300).
* OpenSSH has an option to enable [non-strict acceptor checking](http://www.sxw.org.uk/computing/patches/openssh-patches/strict-acceptor) for Kerberos authentication, so you can ssh to, say, scripts.mit.edu and successfully authenticate despite being load-balanced to a machine that thinks its name is, say, old-faithful.mit.edu. (Specifically a non-strict acceptor lets you authenticate to a machine using any credential in its keytab; a strict acceptor will require that you authenticate to the specific key for the machine's name.) Port the non-strict acceptor option to [Cyrus SASL](http://asg.web.cmu.edu/sasl/) so that scripts.mit.edu can pull the same trick for SVN and LDAP and so forth.
* I often find that [sbuild](http://packages.debian.org/stable/sbuild) installs many of the same packages into the "base" chroot; for instance, a bunch of Debathena packages depend on cdbs and config-package-dev. sbuild should have the ability to take advantage a chroot with these packages preinstalled (so long as all the packages in this chroot still are a subset of the build dependencies), and for extra awesome bonus points, look at a repository and suggest what non-base chroots I should create.
* [Write a caching NSS module](http://debathena.mit.edu/trac/ticket/486) that will play more nicely with Debathena than nscd (the current solution) does. It will probably end up looking like nss_nonlocal.

_Contact: mostly geofft_

## Binary compatibility between OSes

Help the cause of OS ecumenism! FreeBSD provides [binary
compatibility](http://www.freebsd.org/doc/handbook/linuxemu.html) with
Linux operating systems: an add-on module to the kernel knows how to
interpret Linux-"personality" programs, just like the base kernel can
interpret FreeBSD-"personality" ones, and a standard component in the
FreeBSD ports system will install a number of libraries from Fedora 10
in /compat/linux.

However, the number of people who use FreeBSD as their own OS is limited
compared to those who use Linux or Mac OS X. Possible projects here
include BSD binary compatibility on Linux, or porting FreeBSD's Linux
binary compatibility to Mac OS X (which has a bunch of BSD-like parts,
but whose core kernel is actually Mach). The latter would allow running
Linux locker software on [MacAthena](http://macathena.mit.edu).

_Contact: geofft, kaduk_

## Consolidate and Update Athena documentation

Athena documentation these days seems somewhat lacking and/or hard-to-find, which should not be the case. Athena is already foreign enough to incoming students, who are more likely to use their laptops than an Athena machine. Help make it easier by consolidating and updating available Athena documentation so that people know here how to use their Public folder, their Private folder, their dotfiles, and so on.

_Contact: zhangc, kasittig_

## Zephyr Client Hints

Some time ago I wrote [a spec for zephyr client hints](http://geofft.mit.edu/p/zephyr-client-hints.txt), optional extensions that zephyr clients can easily implement to add nifty stuff like typing indicators and [preventing zwgc from starting more than once per user](http://debathena.mit.edu/trac/ticket/206) and such. I got lazy before actually implementing these specs, but I believe they'd be relatively easy extensions to both zwgc and barnowl (in their respective extension languages, even &mdash; no changes needed to core).

_Contact: geofft_


## Project Waverly

[Project Waverly](http://waver.ly/) is attempting to rethink collaborative text editing. We are starting a project to do collaborative text (and other
stuff) editing right -- clean protocol, reasonable algorithms, good web
client, good plugins for desktop text editors, nice API, version control
integration, etc. etc.

_Contact: lizdenys, geofft, maria89_

## Your Project Here

SIPB can help you out in terms of both computing resources and
experienced people to bounce ideas off of. If you've got an idea for
something cool to advance the state of computing at MIT or just
computing in general, [[drop by our office|office]] and say hello.