tag:blogger.com,1999:blog-5246987755651065286.post1393724676108205563..comments2024-02-22T16:15:42.388-08:00Comments on cbloom rants: 09-23-12 - Patches and Deltascbloomhttp://www.blogger.com/profile/10714564834899413045noreply@blogger.comBlogger4125tag:blogger.com,1999:blog-5246987755651065286.post-21460571076181235422012-09-23T19:37:52.032-07:002012-09-23T19:37:52.032-07:00Based on what? He quite explicitly claims that he...Based on what? He quite explicitly claims that he does.<br /><br />[reading Percival's papers]<br /><br />hmmm yeah he doesn't really special case executables to nearly the extent that my algorithm sketch (or Courgette) does. He basically just does matching with some amount of mismatch allowed, so that (hopefully) address changes will show up in just a few mismatch bytes. (and he hopes thos mismatches will be very self-similar and thus compressible)<br /><br />This makes bsdiff interesting for things other than executables, but means its performance on exes is way below what it could be.<br /><br />cbloomhttps://www.blogger.com/profile/10714564834899413045noreply@blogger.comtag:blogger.com,1999:blog-5246987755651065286.post-30001996573546095252012-09-23T18:43:48.068-07:002012-09-23T18:43:48.068-07:00I don't think bsdiff does anything special for...I don't think bsdiff does anything special for executables at all.ryghttps://www.blogger.com/profile/03031635656201499907noreply@blogger.comtag:blogger.com,1999:blog-5246987755651065286.post-62038112843269246512012-09-23T17:40:57.202-07:002012-09-23T17:40:57.202-07:00Cool link on Courgette. It's the next step be...Cool link on Courgette. It's the next step beyond what I describe in terms of mapping absolute addresses.<br /><br />I'm surprised how much better it is that bsdiff; I thought bsdiff already had decent executable code special case handling, but apparently not that much.<br />cbloomhttps://www.blogger.com/profile/10714564834899413045noreply@blogger.comtag:blogger.com,1999:blog-5246987755651065286.post-84541326591514977652012-09-23T15:47:42.006-07:002012-09-23T15:47:42.006-07:00Some Chrome-related links:
There was an attempt t...Some Chrome-related links:<br /><br />There was an attempt to do diff-based HTTP using SDCH, which is basically bsdiff + gzip. I might have mentioned it before...kinda lame. If you're going to come up with a new compression standard, why not just do big-window LZ and just initialize the window. Basically: what you said.<br /><br />Something a bit cooler:<br />http://dev.chromium.org/developers/design-documents/software-updates-courgettewon3dhttps://www.blogger.com/profile/09787472194187459747noreply@blogger.com