Changes between Version 1 and Version 2 of Ticket #9936, comment 83
- Feb 8, 2012, 12:40:22 PM (8 years ago)
v1 v2 7 7 - Ideally we should have neither false positive, nor false negatives; that could be reachable: 8 8 - if browsers would react on ETag like we hoped they would... 9 - if we had a per-file fingerprinting based on content (might be too expensive? see also comment:9 ) 9 - if we had a per-file fingerprinting based on content (might be too expensive? see also comment:9) 10 10 - The 'content' method was a way to get no false negatives, but at the expense of a prohibitive cost at start-up, and also had a high false positive rate (a change to //any// resource would trigger the reload of all the others) 11 11 - The other methods I proposed all have in common to have an even higher false positive rate (it's also a global change which would trigger the reload of all resources, and that change can even have nothing to do with a resource change), combined with a risk of false negatives which entirely depends on the procedure used...