History for OtherSystems
??changed:
-
OtherSystems
In an attempt to benefit from the knowledge and experience accumalated by
other package management and catalog systems, i've looked at some of the
following. additional comments welcome.
"Yellow Dog Updater":http://yup.sf.net
An interesting update client for the yellow dog linux distribution
utilizes rdf for information set. Might be interesting as a basis
for a client implementation.
"Debian's apt":http://www.debian.org
The classic standard for good linux package management. I've taken
debian's definition of depedencies for gideon as described in the
deb control file man page.
"Free BSD Port's":http://www.freebsd.org
I don't have much experience with freebsd port's system. I've used
and it works, probably the best thing you can say about a software
system. The ports/package system basically lays out a virtual filesystem
of packages and distribution files. the packages are binary executables
which can retrieve the distribution file and configure it to the spec
laid out in the package file.
"Gentoo Linux's Portage":http://www.gentoo.org
A relatively new linux distribution with package management based
loosely on a bsd style ports systems with some additional features.
Their package management system Portage is written in python.
The package organization of gentoo and of ports system in general
lends itself to easy replication via rsync of the ports directory.
I feel that this type of system is probably the easiest to mirror.
Portage expands upon the ports system to provide better dependency
resolution and global options, ie if you enable ldap in the global
conf when you install postfix it installs with ldap lookups.
"CPAN":http://www.cpan.org
The competition:). I can't find much info on the actual implementation
of CPAN. if anyone knows of any please add it here or send it to
[email protected]
I found one place where some cpan discussion takes place here .
and found a set of scripts related to the operation of a server here
There are a few minor problems with CPAN, first the organization scheme
seems to require only one author for a package.
i found a little bit info on CPANArchitecture from google.
"PPM Server":http://search.cpan.org/search?dist=PPM
Apparently a copy of activestate's ppm server developed by an
ex-activator. not sure what its legal status is. the implementation
looks fairly clean.
in a nutshell, this is entirely xml based on the osd format. the client
and the server appear to store the xml documents descriptions in an
aggregated document. The server has interfaces exposed via SOAP.
"Py PPM":http://www.activestate.com
Active State's python repository. based primarily off PPMServer.
Some legitimate questions emerge as to why duplicate the efforts of
Active State in creating a python repository. Active State provides
many services and resources to the scripting community. My reasoning,
In a nutshell, it does not behoove the community to be beholden to the
closed source services of a commercial entity with a vested interest
in its own python distribution. for information on pyppm architecture
check out Trent Mick's paper on "Active State":http://www.activestate.com
"OSD":http://www.w3.org/TR/NOTE-OSD
The metadata format which forms the basis of Active State's
distribution. The format was pioneered by Miramba and Microsoft
circa 1997. I know of no other actual implementations other than
Active State's package servers.
There are a number things that can be learned from OSD.
"XSA":http://www.garshol.priv.no/download/xsa/
A kissing cousing metadata format of OSD pioneered by xml pythonista
Lars Marius Garshol. XSA offers alot more contextual information
about a package such as changelogs and author info.
"Source Forge":http://www.sourceforge.net/projects/alexandria-dev
Sadly no updates have been made to the source of this pillar
of opensource software development in quite a while, and the cvs is not
open, nor will it be now that VA Linux is focusing on closed source
development of SF.
There are alot of interesting things that can be found in the
relational schema. The concept of a software trove classification
does offer signifigantly more metadata for searching and organization,
although i found the search capabilities of sourceforge to be suboptimal.
The concept of a software trove seems to also have been taken up by
freshmeat.net
"Fresh Meat":http://www.freshmeat.net
One of the biggest clearing houses of software for an open source
developers. Many ideas can probably be gleaned.
"Open ACS":http://www.openacs.org
This community system has its own package management system
with dependencies and file management.
"Ciphon":http://www.sf.net/projects/ciphon
A python client implementation by Suchandra Thapa
"Swallow":ftp://tummy.org/pub
a catalog server, built with python cgi's and postgres, by
Sean Reifschneider.
Net Distro
Amos Latteier's catalog server based on an early version of Zope's CMF.
currently unavailable..
Other Systems i haven't looked at
- "Red Hat Update Network":http://www.redhat.com