Philip Charles wrote:

> On Fri, 31 May 2002, Patrick Strasser wrote:
> 
> Cut
> 
>>To sum up, first we need a list of files that can be found on mirrors
>>with their corresponding location, what I'm trying to do ATM. And second
>>the files itself for the checksums, which we get from the mounted image.
>>I now work with a local copy of alpha, just to be able to separate the
>>server locations in the jigdo file, but this should not be needed with a
>>file listing from alpha and some file list manipulation.
>>In my theory this would result in the perfect hurd-H4-CD1 jigdo files.
>>(But as a wise man said: I practice the difference between theory and
>>practice is mach bigger than in theory ;-)
>>
> 
> I realy don't like this complexity.  I can add quite a few more
> complications, e.g. unique Hurd boot-floppies which only exist on the
> master CD images.  These would always have to be fetched from somewhere.

Perhaps my explanation was a bit complicated. Let me try to split it in two.
First:
I think for your problem of getting the image as cheap as possible 
around the globe, the solution is already done: You have a new image 
with some special files and and files available around the time of 
shipping. Cut out the fetchable files and ship just the rest. This is 
was jigdo does. You just have to have the packages in a filsystem 
hierachy like on thew mirrors. I gues, tha's what you have annyway.

Usually you need to have a lokal Debain mirror. But creating the jigdo 
files can also be done by specifying a list of files which can be found 
   on a mirror and having just these files local. That's what I want to do.

Actually, the processing of the file list can be done by a quite simple 
script, so creation of jigdo files is completetly automatic.

The second:
If someone would like to download the image via jigdo, you have to 
assure he/she gets all the cut out files. If this files were deleted on 
the mirrors, the ige becomes useless, even disadvantageous, as you have 
to try to fetch the files, to finaly find out som can't be found.
Two ways to solve the problem:
a) A fallback directory for old files. The files get the checksums as 
names. Jigdo can handle this, so this needs nearly no afford at creator 
and user side. You just need a script for storing your old files in the 
fallback directory. And such a script already exists.
b) We can guarantee the images to work until a certain expiration date. 
the files must be on the mirrors until this date, or we don't cut out 
files which are known to change until expiration date. I don't know if 
this is doable.

ATM I'd need the file listings to test my theories. Then we will see 
what can be done with jigdo.

Patrick

-- 
Engineers motto: cheap, good, fast: choose any two
Patrick Strasser <pstrasser at bigfoot dot de>
Student of Telematik, Techn. University Graz, Austria


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]

Reply via email to