Plan for tomorrows (20061213) FESCO meeting

[Michael Schwendt]

On Thu, 14 Dec 2006 11:37:35 -0500, Chuck Anderson wrote:

> Is there a way to run repoview on the normal packages such that it 
> doesn't recreate new files on every run, wasting everyone's bandwidth 
> who has to sync the new files over and over again, even though their 
> contents should be identical to the ones generated before?

Short answer: It doesn't recreate files by default. Since end of October,
any repoview page that is updated contains changes actually.


Long answer: 

Inside the repoview code, it is called "smartWrite" and doesn't rewrite
a html page when its checksum is unchanged, Still, Repoview recreates 
more files than necessary, because adding/removing packages changes the
list of adjacent package names in the left frame of every page.

But to make it worse, old code in the extras-repoview.py script removed
the entire repoview directory prior to running createrepo, because its
existence breaks createrepo <= 0.4.4. So, for a long time, all repoview
files have been created from scratch. It has not been considered a
problem, and I haven't had interest in the repoview script either until
reviewing it begun. By chance, around the same time, Jakub Jelinek
contacted me and told me that the thousands of recreated repoview pages
cause a lot of unnecessary load for mirrors (even with rsync), and I
committed a fix for it on Oct, 28th. Since that day, any file that is
updated by Repoview contains changes actually.