Jeff Spaleta wrote:
Okay ive looked over initially how the check works.
And correct me if I'm wrong but...all you are doing is checking the header.info
on the mirror against the server. This might not add much in terms of
accuracy since its quite possible that a mirror might get the header
before the actual package. So this method of generating a mirror-list
is not going to catch all possible problems. And in fact you might
actually cut out mirrors that have the package but not the header yet,
which means cutting out a mirror that is actually valid for the
downloading step even though the header hasn't been copied over yet. Maybe it would help if there was a re-statement of the problem you
were trying to solve to guage how effective this attempt was at trying
to solve it.
-jef
I'm looking at the errors reported for each mirror and looks like the idea of using the header.info file is not a good idea at all. First , because of the reason you talked about. Second , most of the mirrors which the script said are not ok have the 2/x86_64/os/Fedora/ directory , but they didnt have the header.info file...
At least now we have a script that works. I'll restore the html parsing I've implemented and use it instead of the header.info. And I'll have a check on the man pages for wget . I need to find out which exit codes it uses for errors like file not found , connection timed out , etc , so I can improve the error detection in the script.
-- Pedro Macedo
