On Thu, August 25, 2005 02:40, Steve Palmer wrote: > ... So why not just take this approach: > > - When starting a download, look for the MD5 file in the same > location as the original. It is typically the download filename plus > the .md5 extension. You could check for other extensions/formats too. > - If the checksum file is found, download it along with the main file. > - On completion of the download, verify the main file against the > checksum and warn the user if they don't match. > - If no checksum is found, warn the user that no checksum exists but > still download the file. > - Provide an option in Preferences to disable the warning for folks > who don't really care. Enable it by default but provide a "Don't show > me this again..." option on the warning. > > The advantage of the above are that it isn't intrusive, it is user > friendly and if more browsers support the concept then it should > encourage site owners to always include the MD5 or similar checksum > even on mirrors. If other methods of verifying the file integrity > exist or are developed, they can be subsumed into the implementation > without changing the user experience.
a small note; be aware that this just only checks that the bytes you downloaded match what the server has, IOW that you have an exact replica of the file. It doesn't say anything at all on security, because noone garantees you that the checksum is correct (or equal to the original issuer). Getting a guaranteed correct checksum is impossible to do generic, IMHO. So the only advantage of going through the hassle of doing this is to know even more sure than you know already that your download has completed. Since TCP handles package corruptions and losses, and the server usually tells upfront what the size of the file in bytes is, it is unlikely that you get a file which does not equal the file on the server. Nonetheless the extra check doesn't harm, however, I feel that if this is an emergent area, it should be supported in the protocols (FTP or HTTP) themselves, instead of on top of them, because they are fuzzy checks, and the server treats the checksum just as a file, nothing more, so it can be out of date or anything else. The checksums are *not* related to any files from the server's point of view. _______________________________________________ Camino mailing list [email protected] http://mozdev.org/mailman/listinfo/camino
