Re: Who's working on centralized management?
On 20 Apr 1999, Brederlow wrote:
> Javier Fdz-Sanguino Pen~a <email@example.com> writes:
> > Yes, this was one of the ideas in my own lab. However, since the
> > clients have 6Gb+ of disk I thought it was quite a waste of space to let
> > them install there only NT and reserver a "small" partition to Linux.
> > When the clients might run Netscape/Gimp/Gnome/Wmaker, I think it's
> > necessary to have all binaries local (however a server's /usr is mounted
> > /usr/local). Also, I did not want to break Debian by not letting it touch
> > /usr.
> > I think, however, that such a modified boot-floppy as well as more
> > documentation on how this is down could make a useful debian package :)
> > Regards
> > Javi
> Have a look at coda. Its a network filesystem like nfs or afs, but its
> far better than nfs and more free than afs.
> The main advantage is that it caches data it got from the server. You
> can tell coda what to cache and how long. When you have /usr mounted
> via coda, it will be in cache after a very short time and thus the
> bandwith isn`t wasted.
> Another benefit is that you can have several servers working as backup
> and that you don`t need any server as long as you only use cached
> May the Source be with you.
> PS: Coda is in experimental, but needs recompilation (libc6 / libc6.1
Well, coda is extremely cool theoretically, but the problem is - it is too
far from being ready for production. E.g. ordinary users can easily crash
coda client by attempting to mount already mounted volume.
Another thing: in order to mount /usr via coda, it requires heavy
modifications, since coda itself is in /usr, and this path is hard-wired
into the code AFAIK.