From: "R P Herrold", Saturday, August 22, 2009 8:54 PM > The problem is this -- a vhost.d and linkfarm constellation > works (for some meanings of 'works'), and is not unheard of -- > but it also contemplates adding directories not identifiable > by: > rpm -qf /path/to/vhost.d/templates > > is note integrated with SELinux, and it not accompanied by a > documented or LSB or FHS model management tool (see, eg, > alternatives, or chkconfig) > > Local extensions are all well and good; but the CentOS > approach is conservative, and not developmental; it is about > management within the model of the upstream, of a form that > will not get 'tromped on' by an async upstream security > upgrade, and automatable sysadmin provisioning and management > tools. > > We have the memory of the 'cacheing nameserver' and 'bind' > named.conf changes mid release causing outages upon the > unwary. Those using non-upstream docoed's approaches were > caught when a local extension was stepped on by upstream. > That means we at CentOS, when we extend, package sources into > RPMs, with directories that SELinux is comfortable with, and > use versioned tools so delivered. > > I strongly suspect that the draft model of links needs a raft > of SElinux modifications as well. Haven't tried yet, as > frankly, it strikes me that this type of work needs to be > thrashed out in the Fedora context and rough and tumble of > development. It is just not where the CentOS wiki needs to > be, in my opinion. > > 'wolfy' used the executive sumamry and telegraphic model to > communicate this which we use in IRC when proposals like this > arise; I hope this longer form is not considered 'sarcastic' > > -- Russ herrold Telegraphic. Nice. Hopefully, not. I find link farms annoying in general. Actually, I agree with the conservatism. I'm not suggesting that CentOS change the distribution. I was just suggesting a slightly different method for what many of us were already doing.