Contents
10 year goal
vesta replaces make. vesta is installed on the majority of linux distros and is the defacto standard for building software.
It may take longer.
Intro
See also MarketingStrategy. This page is supposed to describe a progression toward the issues and plan presented in MarketingStrategy which is more complete.
The world changes when you can assume that vesta is installed and running -- ie sw distribution could be done via replication (altho this runs counter to the current bittorrent movement). autoconf would need to be re-written to look at what you have installed in vesta and select the right package version for your system, replicte, build (cross-site cache hit?) and ship (or just run in vesta. Or in reiserFS :)
Replication can not ever realistically compete with BitTorrent. However, since sources are immutable it would be possible to apply BitTorrent's basic method to a re-designed replication protocol.
Identifying goals (e.g. get Linux distributions to include Vesta) and specific tasks we can undertake to reach them is important even w/o timelines.
See other notes below.
10 Year Plan
year |
Nominal Plan |
Alternate Plan 1 |
Comments |
2006 |
major technology company endorses vesta |
||
add to vesta advanced history-aware merge |
|||
2007 |
proliferation thruout major technology company |
|
|
2008 |
due to industry influence, other companies find working with major technology company easier if they use vesta too. |
|
We should encourage vesta users to NOT ship a model and ftp it to others, but rather have other groups install vesta, then email them an immutable version, have them replicate, and build. |
|
This sort of reminds me of PDF or .pps powerpoint files. The "viewer" is free and simple to install and use. The authoring software is more expensive and more complex. Perhaps we can pre-configure a vesta rpm who's goal in life is to install, replicate, build, and ship something. Can we imagine a push-button way to accomplish this? |
||
prepare web site for 1000s of test-drivers. in 1 hour they should be able to dl vesta, install, run demos that show off vesta's key benefits. |
|
||
[sometime in the last 3 years -- windows port. Or if not, do now in 2009 and push back the rest 1 year.] |
|
A Windows port would be really hard, mainly because it lacks anything like chroot so making the build system work for Windows requires developing our own method of intercepting filesystem accesses. I have heard that some people have done things like this, but I don't know any details. We'd really need a motivated Windows developer to do this. We don't have one now, and Intel has no real motivation to fund this development. It's worth noting that some of the original researchers are now at Microsoft research. They may be able to provide some assistance, but they're not going to get it done themselves. |
|
2009 |
now 1000s of vesta users world wide. dozen part-time developers (2x from today?). All due to local commercial decisions. Now it's time to do world-wide marketing. Magazine articles on the woes of make and the fix that vesta has. Conferences. Things like "vesta for cvs users", "vesta for bk users", "vesta for subversion users", etc. Website is prepared. |
|
I think there are some better places than individual application groups to focus where Vesta's advantages would be clearer. How about trying to get an OS vendor to use Vesta to develop their OS? An OS is a huge collection of software. When fundamental pieces are in flux (the compiler, the C run-time) it's important to recompile the right pieces, but it would be best to not waste computes recompiling too much. What if a Linux distribution started using Vesta for development? What if Apple did? |
Tech conferences with flagship open source projects: mozilla/firefox, open office.org, apache, perl, python, gnu, php, morepheus, azureus (or 2009 analog), ad-aware, spybot, vnc, etc. Nice to land a couple of these guys. |
|
||
2010 |
port to any popular linux distro we're not already on; encouraging linux vendors to list vesta as an option on their install site |
|
Don't limit ourselves to Linux. There's a lot of other UNIXes out there. Ideally, I'd like to see Vesta running on all of them. Solaris has shrunk, but not disappeared. The BSDs are not as strong as Linux, but they're quite significant. Mac OS X is pretty significant. Some people (and I'm not saying I agree) would even argue that Linux has peaked and has factors limiting truly wide adoption: http://blogs.zdnet.com/Murphy/index.php?p=434 |
2011 |
encourage world-wide proliferation and consensus that vesta is the new standard |
|
|
2012 |
encourage linux distros to ship and build vesta by default |
|
|
2013 |
continue encouraging |
|
|
2014 |
add world-wide vesta features which assume that vesta is running everywhere. |
|
|
2015 |
declare victory, retire, speak at LUGs around the world about how when i was a kid we had to use 'make' to build. |
|
|
Merge
Consider merging w/ another project, like ResierFS or gnu arch or clearcase or mercurial. The advantages are that it reduces one more competitor, makes more development resources available to the combined effort, and might short-cut our merge development. Or even convince [a notable commercial revctl vendor] that they're going out of business and they can have a successful open source maintenance, support, training business if they open source their code. Then merge w/ [a notable commercial revctl vendor].
Disadvantages/challenges are that A few of the other free VCS projects have a distinct design philosophy which produces a unique user experience (Darcs, monotone). Many others have sprung up over the last 5 years for basically no good reason other than the fact that the developer took issue with some minor detail of existing systems and thought they could build something better themselves. (git is a perfect example, as it's practically a clone of monotone, just optimized for the peculiar case of the development model of the Linux kernel.) It's unlikely we could get such curmudegeonly hackers to be happy working in a Vesta context. The Arch project is struggling (and is considered by many to have non-optimal UI), though there are several derivatives of it (Bazaar, Bazaar-NG, ArX)
Other Notes
While we can ever seriously expect to completely supplant make, our goal should be that Vesta is the defacto standard for building software. make will never entirely go away; there are just too many places where it gets used, even outside software development. (For example, on systems which use NIS to distribute user/group information to multiple clients, the usual way to update the distributed tables from the OS ones is cd /var/yp ; make. Vesta doesn't even make sense in that context.)