[clue-tech] Mirroring Debian testing, i386.

[Angelo Bertolli]

David L. Anselmi wrote:
> Keith Hellman wrote:
>> On Mon, Dec 24, 2007 at 12:58:56PM -0700, David L. Anselmi wrote:
> [...]
>>> So its overhead for checking for changes is wasted.
>>
>> This isn't my experience (I use rsync for all my backup scripts).
>> Granted, I probably don't do the volume of bits or inodes in a debian
>> repo.
> >
>> If what you say is true, then the -I option to rsync doesn't make much
>> sense, and I wonder why it is provided:
>>  -I, --ignore-times
>>    Normally rsync will skip any files that are already the same size and
>>    have the same  modification time-stamp.   This option turns off this
>>    "quick check" behavior, causing all files to be updated.
>
> I didn't mean that the overhead is large.  But it checks time stamps 
> across the network on over 21k files.  That may be required on for 
> backups but not for mirroring (in this case anyway).
>
> apt-mirror downloads the new Packages file and checks the size of all 
> the files listed there against what's on disk.  I think it's 
> impossible for a debian package to change and keep the same size and 
> name (which includes the version number).

I think so too:  any new packages or changes should have an updated 
name.  The optimal way for the script to handle the mirror in this case 
would be to simply generate file lists, and copy only those that don't 
exist.  But I'm really not sure how much of the time it's taking has to 
do with checking the timestamps...

The wording for -I says it causes "all files to be updated" which is 
confusing since it would make more sense if it merely relied on file 
size.  The wording makes it sound like it will just copy all the files.

Angelo