[00:53:24] --- jaltman/FrogsLeap has left: Disconnected [01:06:38] --- Russ has left: Disconnected [01:56:59] --- pod has become available [03:20:24] --- lama has left [04:07:44] --- lama has become available [04:35:00] --- jaltman/FrogsLeap has become available [07:29:00] --- deason has become available [07:31:31] --- lama has left [07:56:27] --- reuteras has left [08:13:36] --- steven.jenkins has left [08:14:01] --- steven.jenkins has become available [08:22:38] (stark) I would like to see Rx and Ubik as separate libraries from anything specific to OpenAFS; if I want to build or use something that uses those libraries and isn't OpenAFS, I should be able to do so. Especially since there can be other AFS implementations based on the same Rx. Probably Rx security classes should be separate libraries from Rx, because they will have class-specific external dependencies. [10:15:11] --- steven.jenkins has left [10:17:18] --- steven.jenkins has become available [10:22:20] --- rra has become available [10:27:24] --- jaltman/FrogsLeap has left: Disconnected [10:38:08] Yeah, I agree that both Rx and Ubik are generic facilities that should be provided as separate libraries since it's reasonable to want to use them without using AFS, and Rx security classes should probably be separate from the core Rx library due to external dependencies. [10:40:34] Beyond that, it's not possible for client and server stubs for a single RPC package to appear in the same program/library unless that package has a server symbol name prefix [10:42:54] It would be nice to fix that somehow. [10:56:54] Uh, impossible? [11:00:07] I'm not sure why. The API that we expose to local programs doesn't have to match the wire names of things, so this is a naming choice in rpcgen and it could make another naming choice. It would, of course, change the historic API to make a different naming choice, but if we're changing all the library names anyway, that's somewhat on the table. Not that I want to gratuitously change function call names for no reason, though, so maybe it's not actually a good idea. [11:00:12] But it doesn't seem like it would be impossible. [11:00:58] Or maybe what you mean is "the way to do that is to add a server name prefix", and hence it's impossible to fix the problem without, er, fixing the problem? [11:06:44] Yes, that's what I mean. I do not consider a server name prefix to be part of the wire protocol, even if we choose to specify wire protocols in the form of .xg files or fragments [11:07:28] Ah, okay, yes. In that case, read mine as "hm, maybe we should add a server name prefix to those cases that are missing one when we see a good opportunity." [11:08:38] Yeah [12:09:55] > CMU using ADM [12:10:12] was that the tool written in scheme? [12:10:22] it's the tool which is scriptable in scheme [12:10:26] it's written in c [12:11:35] ah, ok. [12:11:47] Well, part C and part scheme; the division really isn't where it ought to be. And the scheme is really bad, IIRC [12:13:14] the scheme is a restricted subset of scheme. and when i needed new primitives, several times i had to write them [13:38:50] --- sxw has become available [13:43:40] --- lars.malinowsky has become available [13:51:41] --- sxw has left [14:54:02] --- lars.malinowsky has left [15:30:24] --- jaltman/FrogsLeap has become available [15:42:15] --- deason has left [18:17:22] --- rra has left: Disconnected [18:36:24] --- Russ has become available [21:18:57] --- deason has become available [22:03:15] --- deason has left [22:43:36] --- steven.jenkins has left [22:46:24] --- steven.jenkins has become available [23:02:38] --- reuteras has become available [23:03:29] --- steven.jenkins has left [23:04:08] --- steven.jenkins has become available [23:10:09] --- steven.jenkins has left [23:11:30] --- steven.jenkins has become available