On 11 Apr 2002, Robert Tiberius Johnson <[EMAIL PROTECTED]> wrote: > Thanks for all your hard work on rsync. I think it is a great tool. > I'm especially excited to hear it is used in Intermezzo. I like your > rsync/debian web page.
I'm glad you like them.[0] > I feel you aren't fair to diff (in Section 2.1). For example, I diffed > two Packages files, and then compressed the diff. I also computed the > xdelta of the files. The compressed diff was smaller (by 10%-20%). It's quite true that diff will do much better for some cases, and the Packages file is one of them. I'll update the document to be more fair about that. Of course the most important advantage of rsync is what tridge calls the "morning after pill": you don't have to still have a copy of the old file to be able to generate a diff relative to it. If we accept that it's not reasonable for servers to retain information about every previous version of a package, then something related to rsync is the most likely solution. > Also, in section 3.15, you say, "Fetching a series of deltas which > update the same area is less efficient than calculating the deltas > directly." This is true, but by my computations, the number of packages > which change more than once in a month is 300, which require about 30K > of wasted bandwidth. Sorry, I don't understand what the 30kB number relates to. Is that only for the updates to the Packages file? -- Martin [0] I'm obviously too polite to stay on debian-devel. -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
