El Viernes, 23 de Mayo de 2008, Ciaran McCreesh escribió:
> On Fri, 23 May 2008 09:36:50 +0300
>
> "Alon Bar-Lev" <[EMAIL PROTECTED]> wrote:
> > Or, I guess, you did not managed a large scale complex network.
>
> If you're managing a large scale complex network using quickpkg and a
> package manager on every system, you haven't managed one either.
> You've just mismanaged one very ineptly...
>
> What you're describing is only a sensible solution for up to a dozen
> machines. After that it gets silly.

So, you're imposing to users to use a single network filesystem that
gets as both performance bottleneck and failure point. C00l. And then,
your AFS breaks (because shit happens or because intermediate routers
our switches fail) down and nobody has any access to any software on AFS
outside their local caches, leading to really unpredictable
behaviour. Yet another question, as I don't want to use image-based
management systems, shall I re-compile openAFS packages on each node
(few hundreds), or can I have a way of installing pre-compiled packages?

Binary packages (quickpkg, RPM, whatever) is just an intermediate step
for large-scale package management. A step that Paludis can provide (if
devs want to, of course). The remaining steps are outside Paludis
scope. On the other hand, pre-built images work only on offices that can
afford to shut down their nodes at night, others don't get that lucky.

Please, stop telling users how to run their sites, most of them know
what they do.

_______________________________________________
paludis-user mailing list
[email protected]
http://lists.pioto.org/mailman/listinfo/paludis-user

Reply via email to