[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Error DVD iso jigdo :(

On Friday 24 September 2004 23:30, Richard Atterer wrote:
> On Fri, Sep 24, 2004 at 07:37:20AM +0300, George Danchev wrote:
> > If you use (very) old .jigdo/.template pair(s), and if passing regular
> > debian/ and debian-nonUS/ mirrors to fetch the files from gives you "some
> > packages could not be downloaded" error (because some packages have been
> > dropped in favour of their new versions), then you can use the
> > http://snapshot.debian.net as your debian/ and debian-non-US/ ones with
> > date lines around that .jigdo/.template pair was created.
> > For example, If:
> > zgrep Info sid-sparc-1.jigdo
> > Info='Generated on Thu, 22 Jul 2004 01:38:00 +0200'
> >
> > then your snapshot server lines you can pass to jigdo-lite to fetch the
> > files from should be somewhere around this date, or 2-3 days after this
> > one: http://snapshot.debian.net/archive/2004/07/22/debian/
> > http://snapshot.debian.net/archive/2004/07/22/debian/
> >
> > This sould go to the jigdo documentation at some point since it is very
> > likely newcomers to jigdo not to guess of how to mount an uncompleted
> > iso.tmp file if nesessary and how to find the dropped packages from the
> > snapshot server.
> Loop-mounting old .iso.tmp files is actually documented in the Debian jigdo
> mini HOWTO.

Right. Agreed.

> > if nesessary and how to find the dropped packages from the snapshot
> > server.
> The creators of .jigdo files can simply add snapshot.debian.net to have
> jigdo use that server. So far, I'm a bit reluctant to recommend this,
> because it will mean that snapshot.d.n will be hit badly by people from all
> over the world, when typically these people would be off far better with
> just using the latest weekly .jigdo files.

Yes, I agree that this server is a valuable resource, which must be used as a 
last resort. 
But the idea of  using the latest weekly .jigdo files will still not work in 
the corner cases where there are no a fresher .jigdo/.template pair and the 
user still want to build an image using the old pair pointing to packages 
that have been already dropped. Then, the only solution is to fallback to the 
precious snapshot server (of cource, if the creator of that old pair didn't 
provide any url to a fallback direcroty containing the files that pair was 

> In the long run, my plan is to implement a simple "phone home" system in
> the jigdo GUI app: A CGI script on some Debian server collects the md5sums
> of failed individual downloads. The jigdo application fetches this list of
> md5sums before starting a download, and uploads its own list of failed
> download md5sums to the script after a failure. Furthermore, the downloads
> in the list provided by the CGI are tried first, so if one of them really
> fails, jigdo can abort early with an error, and not after hundreds of MB
> have been downloaded.

Woo, that a bit a complicated effort ;-) Does it worth to be implemented just 
to warn the user at the beginning that he needs to check for a 
newer .jigdo/.template pair ? Again we will go to the above case if there are 
no newer pair(s) for the architecture or the CD set which the user might want 
to construct, and if he still insists to have the image constructed, he will 
complete his "failed to download files" from the snapshot server. 

I guess that this snapshot server is so valuable, that in the short future we 
might need some enhancements to be implemented, like the hierarchical anycast 
technique [1] used for Global Service Distribution. Yes, I know that it is a 
volunteer effort and depends of hardware and human resources.

[1] http://www.isc.org/pubs/tn/?tn=isc-tn-2003-1.html

pub 4096R/0E4BD0AB  2003-03-18  <keyserver.bu.edu ; pgp.mit.edu>
fingerprint 1AE7 7C66 0A26 5BFF DF22 5D55 1C57 0C89 0E4B D0AB 

Reply to: