On 06/03/10 01:32, Pat Villani wrote: > What I have in mind is a design in which all questions are asked up > front and responses are tested for validity, e.g., do you have enough > disk space for all the selected software, do you have the right hardware > for a selected option, etc., and then the install happens. In my > previous experience, we had a GUI that wrote a file to a RAM disk, and > then invoked a process we called playback to to the physical install. > We also wrote this file out to the installed disk so that an installation > could be easily cloned.
I am very keen on installers that allow human interaction to be all over and done with early on, and then you can leave the computer to get on with the copying and so on. I produced a system years ago that did a lot of this, including a system replication facility to format and install DOS plus other partitions (including PS/2, so you can guess how long ago it was!). I also am thinking of writing a new FDISK program (some may recall my Turbo Pascal FDISK from times long ago, that had quite a few command line options and disk tests, and a nice blue interface). The situations I would like to see handled are: 1. simple diskette (or USB or CD or net-boot) of DOS that comes up quickly with a menu to allow the current system to be tested, a new system to be installed, or the option of dropping down to a command prompt. 2. A net-installer for PCs that have odd things happen to them (e.g. used for testing dubious hardware, checking for virus activity, used by students) so everything that was there can be wiped and the disk restored afresh quickly. 3. Easy installation of FreeDOS onto a virtual PC (especially for dosemu within linux, but as general as possible), so a user's partition is mapped and everything is set up idiot-proof and neat from the start. 4. Installation and replication of DOS (and other systems) quickly on a new computer, such as for vendors selling systems without Windows to install at least something that proves the computer works. Especially should consider the case of selling or installing reconditioned PCs where the previous owners' files are sure to be wiped and a simple working system and memory/disk/etc test programs are installed. DOS can still be a good first step in an install process (as an alternative to grub, for instance, yet with the ability to do lots of other things). System cloning tools need not care about what operating systems are being installed, and the code has a large overlap with what FDISK should do anyway, so I am keen to make an FDISK with replication abilities (that might make some of the FreeDOS install tasks easier?). > Additionally, we need to improve software packaging so that any > dependencies are checked prior to attempting installs and packages have > an opportunity to run run pre- and post-install processes or scripts. > It should be a single file, ala rpm, although there are strong arguments > as to why it should be something akin to the windows installer. I like the idea of self-executable zip archives for distributing packages, and some standards for naming documentation (that might make OpenOffice interchange easy) and a policy of a compressed source tree archive within the archive, again: with a standard for naming files to avoid clashes when upgrading, avoid duplication, and make it clear to people looking at the files what they are. Ideally, there should be a ready-to-use executable in the distribution that will work on the lowest common denominator computer, plus it should be possible to take the source tree and produce a new executable optimised for the local hardware automatically by an installer program... which I really would like to see named "Packman" (for package manager). As with rpm and deb packages, there should be a searchable record of files involved with each package, packages installed, and packages available for installation. It should be possible to set policies to only ever use packages from selected sources (already on local disk, from CD/DVD, or selected websites), and another security-conscious requirement: to easily compare the contents of all files with the distribution package (or just a selected file on the PC) and say whether the file has not been changed/corrupted/infected/updated, or (for config files) what changes have been made. Mark Aitchison, Christchurch, New Zealand. ------------------------------------------------------------------------------ Download Intel® Parallel Studio Eval Try the new software tools for yourself. Speed compiling, find bugs proactively, and fine-tune applications for parallel performance. See why Intel Parallel Studio got high marks during beta. http://p.sf.net/sfu/intel-sw-dev _______________________________________________ Freedos-devel mailing list Freedos-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/freedos-devel