Michael Scherer wrote:
> On Monday 11 August 2003 18:27, Buchan Milne wrote:
>> Levi Ramsey wrote:
>> > Another argument in favor of splitting is that it minimizes the
>> > need to upload a massive package when one small component
>> > changes...
>>
>> Uhh, no, in this case the source packages aren't being split, so an
>> update to one small component will mean an upload of a lot of small
>> pacakges, which is actually probably more likely to fail (one package
>> in 10 not making it can cause problems for the other 9 if it's the
>> wrong package) , than the upload of one big package.
> 
> rsync can take this in account and tranfer less bytes, because if 10 
> packages are uploaded, but only one is really a new version, the other 
> would only changed by some headers and the changelog.
> 
> so, this can decrease bandwidth for mirror ( at the cost of having more 
> headers )

As long as you're using rpmsync :o)

(and as long as the compiler hasn't changed)


Reply via email to