> Package Management > *********************** > I was reading through the mail list archives and saw that > package management > is something that is needed/wanted. I agree wholeheartedly. [...] > good fit. Having a > central registry of all these "packages" that we can sync to > would be ideal. > Letting "maintainers" update these packages would be even > better. Has anyone > put specific thought into this? If so, what was it? And, what > would be the > next steps? If we all need it and it is just going to need > someone to write > it I would like to volunteer. My expertise is in web > programming anyway, > that seems a logical way to do it. I have a ton of bandwidth > at work (I am > on a university campus) and could have a test server up nice > a quick. If no > one else has really looked into this I can start to formalize > my thoughts > and present a design recommendation to the list.
If you're thinking of a "Package" being something like "Patch Q123456", or "Adobe Reader 6.0", then you're going to come against problems with redistribution, because unlike Gentoo's portage tree, most of the packages you'll make for Windows don't allow free redistribution of the program, so the best you can do is to provide the installer files for the package, and have the user drop the package files into the portage tree on their local systems. For all that, I think that it's a ripper of an idea to do this, as the biggest problem I've got is making all the shitty little apps in use here install automatically, and I'd *love* it if I could contribute the ones I've done in exchange for the ones that other people have done. Maybe little zip files full of batch scripts and such to install various applications, which people download, stick in their own "ports" tree, and then follow the simple instructions on how to get the program files themselves in there from the CD or original website or whatever. > Post Install Usage > *********************** > Or, Client Package Management. The other side of the coin is > that if our > servers are updated with the latest packages how do we keep the client > machines going. I know that this is a little out of the scope of an > Unattended install project, but like I mentioned above ... Keeping client machines up to date is something I'm wanting to work on (when I find the time). I've played around a bit with using Makefiles for each of the software packages I use, which create stamp files on the client to tell it what the state of the machine is. After (for instance) adding a new version of the software to the Unattended tree, including updating the package's makefile for the new version, a rerun of the master Makefile (customised for the machine's software load, or using database info (see below)) will see the new version in the package makefile and run the appropriate commands to make the update. There's probably a better way (less fragile), but It Seems To Work For Me. > * Administrative interface: Here is an example -- I know I am > having a new > user start, they are going to get the base install, the sales > install, and a > couple of other apps too. I log in to the web interface of my > Unattended > server, select the machine from my inventory, select the os, > select base, > sales, then select the other applications from my catalog of > installs. This Aaah, a man after my own heart. See below. > * License counts: Now I am just wishing aren't I? If I have a > fixed amount > of licenses for a specific application I can decrement that > number when it > is selected from the interface (or maybe when it is > installed), unless the > user is already licensed for it...etc etc You may or may not be aware of the existence of a nice little webapp called IRM. It can do machine, user, licence, networking tracking, (and soon peripherals). My "dream" (called that because I want to do it but have no clear plan on when or how) is to integrate Unattended and IRM, as follows: * Machines are set up in IRM, and are obviously tagged by MAC address and hostname, which the DHCP server uses to assign addresses and whatnot. * Unattended also gets this information out of the IRM database (I'm also thinking of converting a lot of IRM's database backend to LDAP, because LDAP is more suited to the task), so it knows what the machine's hostname and such are. Other random settings can probably be set in the unattend.txt file. * IRM stores information on software installations (including, soon, licence numbers assigned to individual machines), and could have information on where in the Unattended tree the individual software packages are (and the installation script to use). Unattended could use this information to install the software packages assigned to each machine. > If so, what kinds of technologies would you like to see used? IRM is true LAMP. I'm starting to use PgSQL for some of my projects, as the whole relationships thing (as well as triggers and stored procs) is starting to become necessary for my work, but MySQL works nicely for most purposes. - Matt ------------------------------------------------------- This SF.Net email is sponsored by: IBM Linux Tutorials Free Linux tutorial presented by Daniel Robbins, President and CEO of GenToo technologies. Learn everything from fundamentals to system administration.http://ads.osdn.com/?ad_id=1470&alloc_id=3638&op=click _______________________________________________ unattended-info mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/unattended-info