Re: Enlighten me about jigdo ..
On Fri, Nov 14, 2003 at 02:05:03PM +0200, Delian Krustev wrote:
> First are You sure the exact reconstruction of the iso is not possible with
> mkisofs ? What are your concerns, different versions of mkisofs producing
> different isoimages (e.g. this iso made with mkisofs-20031114cvs in the
> image itself) ? Have You consulted the source or contacted the mkisofs
> about that ?
One issue that springs to mind is that IIRC to make one certain
architecture's CDs bootable, the Debian CD scripts actually execute a "dd
if=bootblock of=image.iso".
> Second, if first could not be achieved. Why is it an "absolute requirement" ?
- People like to check the images' md5sum, either to verify it was
downloaded correctly, or for security reasons.
- Other DDs who are concerned with certain aspects of the CD generation
process won't be too excited about any problems introduced by slightly
different CD images. With jigdo, I'm in the nice position of being able to
blame other people if e.g. CDs don't boot. ;)
> I personally don't mind isos with differend md5s if the files inside it
> might be verified and proved authentic. I see no problem in adding
> an option(if it doesn't exists) in the debian installer "verify all
> installed packages with respect to CDROOT/md5sums".
Isn't support for signed .debs already implemented in recent dpkg versions?
> Note this is only necessary for persons obtaining those images and
> not making them with mkisofs. If you keep your mirror uptodate over rsync
> and running mkisofs personnaly the md5s of the isos will not be a concern.
You should also be aware that your solution won't be very popular if it
requires people to permanently dedicate several GB of disc space to a
partial Debian mirror. With jigdo, people can just toss their old CDs to
the program and say "reuse any files from these old CDs which also appear
on the new ones".
Cheers,
Richard
--
__ _
|_) /| Richard Atterer | GnuPG key:
| \/¯| http://atterer.net | 0x888354F7
¯ '` ¯
Reply to: