On Tue, Feb 10, 2004 at 07:41:12PM -0200, Alexandre Oliva wrote: > On Feb 10, 2004, Warren Togami <wtogami redhat com> wrote: > > > Far fewer files are downloaded for headers, meaning more efficient > > data compression and faster header downloading. > > Unless I misunderstand, the one gotcha is the common case of a small > changes in a large repositories requiring the entire set of (single > file) header information to be downloaded. This *could* become a > bandwidth problem, not only for people in the wrong ends of small > pipes, but also for servers, since rhn-applet would probably keep on > fetching it over and over (at least every time it changes). This could be accomplished by an incremental approach. In a two file scenario there could be a weekly updated base file and file with up to date differences to that. Every week the latter file gets merged to the weekly file and is resetted to 0 length. That way the large bandwidth consumption happens only once a week and you have only two files to check for downloading. Add more tiers for adjusting the development/update pace. I only don't know how well the xml metadata diff file could remove entries from the base file (or the higher tier), but I think it would be doable. -- Axel Thimm physik fu-berlin de
Attachment:
pgp00080.pgp
Description: PGP signature