Re: [gulus-miroir] mirror script update / autoupdate feature?
Joerg Jaspert un jour écrivit:
Hi
Beside that, something I wanted to discuss with other mirror admins
first:
I am constantly changing and improving the scriptset. And certain
updates should reach all mirrors, as they have some nice and important
changes. But its a bit bad to always mail mirrors-announce, wait, ping
mirror admins that missed the mail / had been on holiday, etc.
Well, am I the only one that find it a little crazy to use a script
that is constantly changing on a production server? For me, it is a
reason for *not* using it.
When you have a version that you will plan to keep unmodified for at
least 6 months, please tell us. I will look at it carefuly before
upgrading, test it and checks my Debian rsync logs for the next few days
to be sure that the updates still goes fine.
How about an auto-update feature? That, at the end of the mirror run,
checks if a newer version exists and then installs it.
If I was running Fedora Rawhide, I would say maybe, but for most
mirrors, asking them to upgrade their scripts more than twice a year is
probably too much.
If it is really needed, then there is probably something wrong.
Also, sooner or later, there will be an update that will breaks
synchronisation somewhere, or use an rsync option not available for
someone not using Debian (I do, but I also ran OpenBSD for a while).
A better approache would be to have 3 scripts: one stable, one testing
and an experimental one that move to testing after few days if there was
no problems (reminds you something?)
> People should install procmail! :)
I know this is a joke, but I want to make it clear that there is no way
that I will install procmail on my mirror. Or Xorg libraries, before you ask.
I also have an idea that would allow very frequent updates for minimal
load on the mirrors. It is not perfect, but would works with any rsync
versions, and would be backward compatible for the mirrors and the users
if they don't take advantage of it. I will post about it when I have time
to explain it clearly.
Simon Valiquette
http://gulus.USherbrooke.ca
Reply to: