On Tue, Oct 14, 2003 at 09:53:31AM -0400, seth vidal wrote: > > Create in memory or on a temporary file, compare content, then discard > > or actually update. Seems you're recreating the full header directory from > > scratch, okay that works too. > > so I should open the current header(s) and compare the content? Why do 2 > reads? I can just overwrite and the ones that are the same stay the > same. That works too as long as the timestamp is preserved and update is atomic. Now between 2 read and one read and one write (because most of the time from cron you would not write again) the performances can be argued ( if you use NFS changing inodes is really not a friendly thing to do :-), I was really not thinking about performances but about cache, if we start deploying yum in the large scale keeping the cache happy is quite important, I noticed that yum-arch regenerates everything so, I'm wondering about this. To me this is especially important for the header.info files, I just don't want to kill the servers once I integrate Yum support in the rhn_applet and breaking the HTTP cache then means paying the full cost for a new transfer instead of a 304 without content, i.e. you loose at least one order of magnitude in scalability. Daniel -- Daniel Veillard | Red Hat Network https://rhn.redhat.com/ veillard@xxxxxxxxxx | libxml GNOME XML XSLT toolkit http://xmlsoft.org/ http://veillard.com/ | Rpmfind RPM search engine http://rpmfind.net/
