Frans Pop wrote: > [ Snip some excellent points ] > > Downside of this all is that you _would_ IMO need to test the mirror and > possibly that is exactly what the user wanted to avoid in the first > place Yes, this was what prompted me to originally file the BR - a friend had installed via a DVD and it wasn't immediately obvious to him how to add a another mirror. Of course, this info is in sources.list(5) and all over the internet, but it would have been a quicker and less typing if there was something resembling an example in the file. If a working network is not actually part of this use case then, yes, we must conclude from that auto-detection is unworkable. If a working network is, then we could start to look at CDN and possibly autodetection. The latter does sound a little fragile, though. > If you don't want to test, the defaults really should be totally generic > (i.e. _not_ have a specific country code at all, maybe just literally > 'ftp.xx.d.o') and have a lot more explanation instead. I agree. I'm starting to think that maybe this bug can be fixed with just adding: * An example of a "deb http" line, commented out * A short description, highlighting that not all mirrors contain all archs * A link to the list of mirrors In fact, using a vendor-neutral hostname instead of "d.o" such as "ftp.xx.hostname" would prevent the derivative problem to some degree, but I'm not entirely sure how d-i accomodates derivatives (and they may wish to change the wording and links of the mirror list anyway). Regards, -- Chris Lamb, UK chris@chris-lamb.co.uk GPG: 0x634F9A20
Attachment:
signature.asc
Description: PGP signature