Vasiliy wrote: ... > > So we see (me and Rich McAlister) how this request can be done: > > 1. We introduce new type of packages which may be pkgadd by any > customer in their own home directory (some keyword in pkginfo set by > package developers), This packages may be installed by regular user > for testing-evaluation. However sysadmin also able to install this > package on the system if it pass test-evaluation if needed. > > 2. Each user will have his/her own registry - contents file and > var/sadm/pkg where all his/her software will be registered and which > pkgadd will refer to in addition to system contents file. > > 3. Pkgadd and patchadd on system level will not check user contents > file and do not care if software installed by user will be broken > (and it affects only user - no big deal, same commitment level we now > have for tar-balls) > > 4. Pkgadd on user level will check system as well as user software > and will not install software if it is incompatible or require some > other software to be installed. Here we raizing our commitment level > in comparison to tar-ball commitment level. > > 5. Patchadd will check system and user software but will not applay > patches to nonuser software but instead check if it is up to required > level. This way if sysadmin install already same patch it will just > update user software as it doing this right now for partially > installed patches, but is patch not yet installed then it refuse to > update only user software and make it inconsistent this way. >
I'm not seeing how this is any different than the basic proposal outline that James presented to start this discussion. It doesn't answer any of my concerns about presenting a more comprehensive view of the system software, though. Dave
