Re: Building Woody-CDs
On Tue, 15 May 2001, Raphael Hertzog wrote:
> Le Tue, May 15, 2001 at 01:23:53AM +0200, Nikki Claus écrivait:
> > please excuse me for dropping in like this. i am reading some debian
> > groups via web since i installed debian over my suse (debian rulez
> > :))))) this is my first post.
>
> I don't wanna flame you. But still, there's nothing worst than starting to
> rehash something that has already been discussed ... your post doesn't
> (and the previous post where I replied in an inflammatory way) help us at
> all. No progress has been made. No new solutions, only critics. While
> critics are good, there's always one time where we have to take a
> decision. And in this case, the decision has been taken, we won't go
> backwards unless you can provide me a better solution. But it serves
> nothing to say that the current solution is not good enough. Find a better
> one that will meet everybody expectations (and the one that you give below
> doesn't meet Jason Gunthorpe's expectation for apt) before.
>
> > http://lists.debian.org/debian-cd-0105/msg00040.html as i see it this
> > works with all versions of apt. example patch to debian-cd is included
> > and also for apt. why not do it that way??????
>
> Because it's a stupid overhead for apt. Because Jason doesn't like it
> either.
As I explained in that very post, I strongly expect Jason did not (want to?)
understand my proposal. And maybe you don't either. The overhead for APT is:
- when using a full mirror:
- one failing stat() or open() call when using local archive
- one failing HTTP/FTP request (<1k traffic) when using FTP/HTTP
(and the very fact that it's failing is the information you need)
- when using partial mirror (incl CDs)
- the Packages.complete must be read ALWAYS because that's what's signed
- read one Packages file (probably relatively small) in my proposal
versus
stat/HEAD-request 6000+ times (!!) in all other proposals
I'd hardly call my proposal creating a "stupid overhead".
Regards,
Anne Bezemer
Reply to: