[01:03:42] --- dev-zero@jabber.org has become available [01:03:44] --- dev-zero@jabber.org has left: offline [02:28:34] --- sxw has become available [03:10:26] --- sxw has left [03:19:38] --- sxw has become available [05:22:09] --- SecureEndpoints has left: Replaced by new connection [06:10:54] --- SecureEndpoints has become available [06:13:49] --- mmeffie has become available [06:14:37] congratulations on your new role sxw. wishing you much success. [06:15:03] Not so much with the new. We discussed it last year, and it just never got done ... [06:15:21] But yes, thanks. May I spend many years never having to do anything. [06:27:18] --- mmeffie has left [07:02:45] --- matt has become available [07:48:47] And idea occurs, too late for this year, that interfacing OpenAFS with spotlight might be a very interesting project. [07:49:31] never too late. deadline is not until tomorrow :) [07:50:51] Do you want to stick it on the page? [07:51:04] At least then it's there if we're looking for ideas in the future. [07:51:12] send me text [07:59:42] of course I suspect that this is a project that would require access to apple internals to implement [07:59:57] the plugin api for spotlight is for evaluation of document types. [08:00:01] I'm not sure, I'm looking at the Spotlight API documentation to see what's possible. [08:00:15] One of the APIs is for metadata import. There are other bits to it. [08:02:01] the metadata importers are the per-filetype plugins that provide filetype aware context [08:02:24] Yeh. I know. They're not what we want. [08:03:00] the ability to access indexing info from another Mac appears to be an embedded protocol permitting remote spotlight queries. [08:03:16] And we definitely can't have that. [08:04:19] the question is "how is that implemented?" is it an ioctl against the file system? If so, we could implement it and pass off the query to the metadata server that RedBear's project will create [08:04:56] Or is it something that explicitly knows about AFP, and does it as a filesystem operation. [08:05:06] ie, does AFP have a 'do this search' RPC? [08:05:48] But whatever the mechanism is, either to add new files, or do hook searches, Apple don't seem to document it. [08:06:00] no but we can ask. [08:06:21] You probably have a better set of routes to ask with than I do ... [08:08:10] spotlight ignores /afs i suppose? [08:08:38] I sincerely hope so [08:10:11] hack: a dedicated machine that has an /afs which only contains whatever you want indexed. then do remote spotlight queries against that [08:11:03] not that it would map it to the "local" /afs anyway [08:11:24] :/ [08:23:56] interfacing openafs with spotlight has some perils unless you get out of band help from the server, i think [08:25:30] ah, i am behind, you guys already covered it [08:26:26] --- sxw has left [08:34:32] --- sxw has become available [08:35:36] I think the issue with integrating this with RedBear's work, is that what Spotlight considers Metadata, and what he considers metadata are two different things. [08:36:20] Also, in a world where the fileserver is a Unix machine, and you're generating all of the data on your Mac, there's a distinct bonus if the local importers can be used, so that the application producing the data is the one responsible for indexing it. [08:36:45] Maybe Apple is the wrong platform to try and do this on, and we should have a look at what Beagle, for example, offers. [08:47:08] Not sure if there are any prospective students here, but there's not going to be any extension to tomorrow evenings Summer of Code deadline. So if you're thinking of submitting, and haven't done so yet, do it *NOW*. [08:48:57] --- reuteras has left [09:15:58] --- sxw has left [09:40:45] --- Russ has become available [09:43:55] --- edgester has become available [09:45:04] Hi Simon, congrats/sympathies on being the new security officer! [09:45:28] Yeh. Thanks, I think. [09:52:26] do you at least get a truncheon to go along with the office? [09:53:01] Some kind of remote, time-travelling, LART would be useful. [09:53:26] also, the phrase "OpenAFS Sargent at Arms" is now going through my head. [09:54:49] Well, the IETF has one. Fortunately, our requirements to keep order on our mailing lists are somewhat less. [10:10:56] the TSM stuff recently mentioned on openafs-info... is that in parallel to the API you folk were working on? [10:12:23] --- sxw has become available [10:14:34] Pterydactl dudes talked about implementing NDMP...? [10:14:52] Did they submit an implementation? [10:14:59] NDMP? [10:15:26] It's a proposed standard backup arch. and protocol (ndmp.org) [10:16:50] I don't believe Teradactyl did an implementation. They don't really have an incentive to do so, given that what I and probably most other people want it for is to use a backup product other than theirs. [10:17:03] It really doesn't make any business sense for them to implement it, although of course I'd love it if they did. [10:17:54] NDMP is the way that you do backups of storage appliances these days. [10:17:58] It's what Netapp, etc. supports. [10:18:21] I think Kris's talk was more "here's a way you could do ...", rather than detailing what they're doing. [10:18:40] I was really hoping I'd get some money to fund work on better AFS backups, but sadly not. [10:18:47] If AFS supported NDMP, then to our backup system it would look just like a Netapp or EMC box, which would make our storage people very happy. [10:19:22] And as I understand how the protocol works, AFS can store ACLs and whatnot as NDMP objects that are opaque to the NDMP implementation but that AFS can then restore. [10:19:23] I think it would make a lot of people very happy, as would better support in Bacula for AFS. [10:19:33] That's my understanding from what Kris said. [10:19:58] It's on our network file system strategy to fund it. We'll see if that survives the prioritization round. [10:20:33] --- edgester has left [10:21:19] Of course, what our strorage people would like even more would be for the AFS fileserver to understand Netapp file system snapshots so that they can do disk-to-disk replication and snapshots instead of using any dedicated backup system at all. [10:21:35] That seems like it would be harder. [10:22:21] Bacula, doesn't itself implement NDMP, does it? [10:22:24] Especially if NetApp don't tell you their snapshot format. [10:23:00] Not as far as I'm aware. Bacula is slightly orthoganal to this conversation, other than it would provide a great way for sites who don't already have a backup investment to backup AFS. [10:23:02] I google some people wishing it did so... [10:23:06] Well, they expose it as a file system. So basically it requires the file server understand how to read an older copy of its underlying file system and treat it like additional volume replicas. I have no idea how one would even start doing that, but I can sort of see how it would be theoretically possible. [10:23:53] Does it need to do more than just export .backup volumes as a filesystem that the backup system can understand? [10:23:56] Yes, Bacula would be a good place to start one way or another. [10:24:02] Do you need to be able to write to it, too? [10:24:54] You don't need to be able to write to it, but you really want it to go the other way around -- the AFS file server should ideally mount all of the available Netapp snapshots and expose the contents of them as volume.2009-03-20, volume.2009-03-19, volume.2009-03-18, etc., however far back you have snapshots. [10:25:40] In the Netapp world, that's how you do user-initiated restores: the user goes into the snapshot tree and copies out the file they want from the date they want. [10:27:29] That sounds a _bit_ like what Umich would dearly like to be able to do with shadows. [10:27:44] We'd quite like that too, actually. [10:29:58] --- sxw has left [10:34:00] i think teradactyl was shopping ndmp to see if anyone bit. which is fine, i'd have been happy if someone did [10:34:26] --- sxw has become available [10:35:46] --- SecureEndpoints has left [10:35:46] --- SecureEndpoints has become available [10:36:45] I believe what NetApp implements is the Microsoft name space extensions filename:[timestamp][:stream] [10:37:24] the server responds with the best version of the file for the specified date [10:39:35] I wouldn't implement this in AFS at the volume layer. I think a read/write volume that supports up to N versions per file as bounded by quota would be the right approach. [10:41:13] regarding teradactyl and ndmp. the chain of events was: 1. stanford asked me about ndmp backup options. 2. teradactyl was looking for a topic to present at the workshop at SLAC. 3. teradactyl was hoping to get a development contract from Stanford to implement the NDMP support. [10:42:03] Ah. Okay. [10:42:33] I think, in terms of the file:date stuff, it depends on what you want to do. All we'd really like is multiple, off-site, backup volumes. [10:42:53] We get a single backup volume by abusing read onlys. But, please sir, we'd like some more. [10:43:29] fwiw, I know someone who might be interested in doing some GSOC work on backups. He's supposed to call me tonight, and I told him to stop by here. [10:43:44] or, rather, I told him to stop by irc..not here. [10:44:14] that could be done with an extension to the mount point. cell%volume.backup:relative-time [10:44:51] That's an even better idea, yeah. [10:45:13] who needs .backup if I can read arbitrary past versions? [10:47:42] --- sxw has left [11:30:54] --- sxw has become available [11:42:31] --- sxw has left [11:42:47] --- sxw has become available [12:07:28] --- sxw has left [12:12:02] --- dev-zero@jabber.org has become available [12:12:06] --- dev-zero@jabber.org has left: offline [12:22:23] --- mmeffie has become available [13:25:01] > fwiw, I know someone who might be interested in doing some GSOC work have him apply now, even if he may not follow through! [13:37:54] yes, he's to call me when he gets out of class. I told him to stop by the irc channel to introduce himself, too. [16:11:09] --- matt has left [16:15:24] wait, you have college students call you? [16:15:39] what, is this 1993? >_> [17:45:05] --- Russ has left: Disconnected [17:52:19] --- Russ has become available [17:54:08] --- matt has become available [17:54:20] is tkeiser? [18:14:24] --- dev-zero@jabber.org has become available [18:32:15] --- Russ has left: Disconnected [18:48:24] --- Russ has become available [19:22:29] fwiw, I talked with Bill for half an hour or so. We talked about the various suggested projects, and he's going to think on them, look at them some more, and possibly show up on irc later tonight/tomorrow am. I'm not sure what his irc name is, though. [22:01:02] --- dev-zero@jabber.org has left: offline [22:48:31] --- matt has left [23:05:36] --- reuteras has become available [23:32:33] --- Russ has left: Disconnected