20 Jun 2012 06:29
Re: metalink support
Anthony Bryan <anthonybryan <at> gmail.com>
2012-06-20 04:29:52 GMT
2012-06-20 04:29:52 GMT
(moved over from libcurl, since this list is probably more appropriate) > Am 19.06.2012 18:16, schrieb Tatsuhiro Tsujikawa: > > On Tue, Jun 19, 2012 at 5:45 AM, Guenter<lists_at_gknw.net> wrote: > >> Hi all, > >> I did build curl with metalink support, but a test on NetWare failed due to > >> illegal filename; I tested a link from the curl download page: > >> http://curl.haxx.se/metalink.cgi?curl=win64-ssl-sspi > >> this results in the metalink file 'metalink.cgi?curl=win64-ssl-sspi' which > >> is an illegal one since '?' is not allowed on NetWare, and neither on > >> Windows; so we need to think about what we do here: just cut off the > >> filename string beginning with the '?' (simplest) or replace all possible > >> illegal filename characters with an underscore? > >> Or, maybe even store it as a temp file, and remove it afterwards if the > >> metalink download finished and checksum check was successful? > >> > > > > Another solution is use -o option and specify safe name. > > Since metalink enabled curl parses metalink file based on the > > content-type the remote server returned, file name does not need to be > > remote name. I'm sorry if I missed the point.. > ah, ok, I used -O so it was named after the url (though anyway I thought > that some time back - before metalink support - we discussed this issue > and agreed that we cut off the url beginning with '?' for using as > filename); > but then question is: why do we need an download option at all (-o | -O) > ? Whouldnt it make sense to support metalink download also if no > download option is provided but the server headers identify as metalink > in content-type? Then the metalink data could be stored into a temporary > file, and deleted after successful download and checksum verification ...(Continue reading)
> we are already having a similar discussion...if you are using a
> metalink download, by default it will write to the filename taken from
> inside the metalink (except for some sanitizing done within
> libmetalink).
yes, thats clear to me, but this is the filename of the final target the
metalink points to, but I was talking about the storage of the metalink
data itself ...
usually you might find a metalink URL like this:
RSS Feed