Ok, I did some pause to see if anybody try it. Please, let me know if you did, I hope it just work well and this is why there are no replays...
Patching over http-protocol. First let me note that patchadd already support http and https. You already may say -M http://<URL> and then patch_IDs. It was not supported for zones (this project fixing it as a side effect), but otherwise if you have webserver where patches are posted in a JAR-form - then patchadd will get them and install on your system. So I am not introducing http-patching it is already in place. I just make recursive patching (over zones) work with it. Major difference between directory and web server is that I can not lstat it - nor reasonable way in general to get all contents from this location. This make recursive patching same way it done for directories impossible. If I have patch A declared as required by patch B and patch A is not present in specified location because some higher revision of it is obsoleted by some patch it is absolutely impossible to find this out having just name of required patch. If I check location - there is no A there, it is only metadata inside some patch has information that A was actually obsoleted, but to get it I need to know that patch name and I can not check all patches from this location because it is webserver with directory access possible closed or not supported. The way to resolve it is to put all dependency data in one place so patchadd will get it from there and analyze it to trace down dependency - manifest file located on webserver. Format of patch records this file is exactly same as format of patchadd records in pkginfo plus list of packages affected by this patch. To make recursive patching works over http manifest file should be posted on webserver in the same directory where all JAR patches located and them patchadd will first read this file to understand what patchbase is. Thanks, Vassili. This message posted from opensolaris.org
