[00:38:46] --- dev-zero@jabber.org has become available [00:39:31] --- dev-zero@jabber.org has left [02:20:46] --- Simon Wilkinson has become available [02:42:59] --- Simon Wilkinson has left [02:44:03] --- Simon Wilkinson has become available [02:55:28] --- Simon Wilkinson has left [02:56:13] --- Simon Wilkinson has become available [03:18:05] --- Simon Wilkinson has left [03:30:08] --- dev-zero@jabber.org has become available [03:39:58] --- dev-zero@jabber.org has left [03:50:54] --- Simon Wilkinson has become available [03:52:14] --- Simon Wilkinson has left [03:53:24] --- Simon Wilkinson has become available [03:57:46] --- Simon Wilkinson has left [04:35:07] --- Simon Wilkinson has become available [04:44:14] --- dev-zero@jabber.org has become available [04:45:00] --- dev-zero@jabber.org has left [04:45:03] --- dev-zero@jabber.org has become available [05:02:30] --- dev-zero@jabber.org has left [05:28:17] --- Simon Wilkinson has left [05:40:04] --- Simon Wilkinson has become available [05:44:44] --- Simon Wilkinson has left [05:51:09] --- Simon Wilkinson has become available [05:55:08] --- Simon Wilkinson has left [05:56:41] --- Simon Wilkinson has become available [06:14:05] --- Brandon Allbery has left [06:14:26] --- Brandon Allbery has become available [06:16:50] --- Brandon Allbery has left [06:17:25] --- Brandon Allbery has become available [06:17:50] --- Brandon Allbery has left [06:19:23] --- Brandon Allbery has become available [06:20:44] --- Brandon Allbery has left [06:21:00] --- Brandon Allbery has become available [06:24:31] --- Simon Wilkinson has left [06:37:33] --- Simon Wilkinson has become available [06:39:18] --- Simon Wilkinson has left [06:42:31] --- Simon Wilkinson has become available [06:42:33] --- Brandon Allbery has left [06:43:07] --- Brandon Allbery has become available [07:02:33] --- Brandon Allbery has left [07:02:49] --- Brandon Allbery has become available [07:02:59] --- Simon Wilkinson has left [07:12:30] --- meffie has become available [07:20:45] --- Simon Wilkinson has become available [07:34:39] --- deason has become available [08:10:23] --- Simon Wilkinson has left [08:14:49] --- Simon Wilkinson has become available [09:03:14] --- meffie has left [09:15:09] --- Brandon Allbery has left [09:15:38] --- Brandon Allbery has become available [09:30:32] --- m.meffie has become available [09:35:00] --- Brandon Allbery has left [09:35:19] --- Brandon Allbery has become available [09:45:25] --- m.meffie has left [09:47:41] --- Simon Wilkinson has left [09:55:07] --- Brandon Allbery has left [10:15:32] --- Brandon Allbery has become available [10:27:22] --- Simon Wilkinson has become available [10:35:15] --- Brandon Allbery has left [10:52:50] Is there a quick pointer or search keyword for those 2007 discussions of prdb updates? [10:53:19] I suspect they were on snip snap [10:53:35] What's that? [10:53:52] The Swedish AFS-SIG wiki [10:54:11] Ah. [10:54:42] It's where all of the original rxgk design documents lived, too. [10:54:48] --- m.meffie has become available [10:54:54] --- m.meffie has left [10:54:55] --- Brandon Allbery has become available [10:54:59] Is it no longer with us? [10:55:06] So if you've got those, you've probably got the prdb stuff too. [10:55:11] No, the site disappeared a while back. [10:55:58] --- meffie has become available [10:58:04] It was afsig.se [10:58:26] I found a reference to that in the openafs-devel archives [11:06:35] --- meffie has left [11:07:42] If the site hosting design documents is gone, does that mean we have to reconstruct the design from scratch? [11:08:23] I believe that there are copies of some of the site. I've got a copy of all of the rxgk stuff somewhere, for example. [11:08:59] But from what Derrick said, I think the original prdb design was somewhat flawed anyway, as it could lead to you overflowing your pt database. [11:09:49] Is a Ubik extension to allow larger (virtual?) databases is out of scope? [11:10:37] There's a few things that have been talked about there - you can change the maximum size of the Ubik database by going from 32bit to 64bit values. [11:10:53] You can split your Ubik database across multiple files. [11:11:04] Both of those did occur to me :) [11:12:01] Multiple files has been talked about a fair bit. I think the conclusion was that you can't do it within the current RPCs, even though it looks like you should be able to, because there's no way of ensuring cross file consistency. [11:12:02] It's unclear whether we want to get into something as complicated as a DB structure that is not just an array of fixed-size records. [11:12:14] Well, you already have that. [11:12:49] A little - you've got hash tables and things to contend with in the current pt database. And then you've got the joys of supergroups. [11:13:04] Sure. [11:13:04] It's a pretty tangled mess in there right now, even before something else gets bolted on. [11:13:13] I was thinking more about the large space usage front. [11:13:28] Not needing a large fixed-size record for each user could help some. [11:13:59] But given the complexity, it's not clear that it's worth it. [11:14:03] You can't use fixed size records. Exported names can potentially be massive. [11:14:42] I _think_ the main problem is one of fragmentation. [11:14:50] How so? [11:16:30] I'd need to find the original document to refresh my memory, but I think it required certain things to be contiguous. And so with a particularly pathological createname/deletename pattern, you could rapidly use up space in the database. [11:16:59] I could see that. [11:17:41] It's not clear to me that just allowing larger databases is a great way to fix that, because a) It delays the problem and b) It means that you end up massive initial propogation delays because ubik doesn't do sparse files. [11:19:51] A prfsck utility/functionality might be better. But I think I would want to do that offline and then just propagate it to the other sites... [11:20:09] (to repack things, I guess more defragment than fsck) [11:20:25] Well, ideally you want something that's self healing, so it require no administrator intervention at all. [11:20:56] But then it has to work online. Which is possible, sure, but bugs cause big problems. [11:26:01] Or you just have a storage scheme that chains entries using something other than their adjacency. [11:27:24] Which is something of a more radical database format redesign, no? I think I would like to see what the 2007 design was, just for kicks. Maybe I should ask the list if anyone has a copy. [11:28:59] I may do, somewhere. Not hunted properly yet. Derrick might, jhutz might. Not sure who else was there. [11:40:03] --- Simon Wilkinson has left [11:58:40] I swear I checked the wayback machine and got nothing [11:59:31] :) [12:00:53] although finding that was somewhat interesting, I think I didn't get an identifiable link to it until after I followed the link for the ubik extensions (which, at least as of 2007-Dec, were not very well fleshed out) [12:03:06] btw, I actually took that one from 2007-Dec, so it had been unchanged up until then. feel free to scan forward in the timeline for more recent versions [12:04:12] This is what I got sent to when I stripped a date entry. Hmm, javascript is disabled, that might be unhelpful. [12:04:42] Yeah, there's only that one capture. [12:46:57] --- Simon Wilkinson has become available [13:03:31] --- Simon Wilkinson has left [13:03:49] --- Simon Wilkinson has become available [13:47:59] --- Simon Wilkinson has left [13:48:37] --- Simon Wilkinson has become available [13:54:36] --- abo has left [14:01:29] --- Simon Wilkinson has left [14:03:18] --- Simon Wilkinson has become available [14:37:25] --- Simon Wilkinson has left [14:39:43] --- Marc Dionne has left [15:45:30] --- deason has left [15:46:20] --- Marc Dionne has become available [16:07:52] --- abo has become available [18:30:48] --- Marc Dionne has left [21:23:04] --- jaltman/FrogsLeap has left: Disconnected [21:26:06] --- jaltman/FrogsLeap has become available