XDMCP again
Mar. 16th, 2009 09:47 pmWhy does XWindows "documentation" have to be so contorted and tangled anyway? I suspect it's because even the maintainers of the standards and the code don't really understand them.
I'm as far as getting an actual error message now. You have to begin by understanding that in the X world, "server" and "client" are reversed from what you normally think of. So the place where the applications and data reside is called the "client" and your X-terminal is the "server". While it does make a kind of twisted sense, it just confuses every discussion of the subject, and especially so since the writers themselves mix the two ends of the conversation up occasionally.
So the problem appears to be one of "authentication", which is done with what are called "MIT-MAGIC-COOKIES". I've finally learned how this works under circumstances where the user is already logged in at the server end. The authentication cookies are stored in a particular file in the user directory. But when you are using XDMCP to start a connection, no one is logged in yet. So where do the cookies come from and where are they stored? The documentation is of course vague on this, mostly handwaving. But if the two ends of the connection don't agree on the same cookie, they refuse to talk. When one end is Linux and the other is OpenVMS, it gets even messier trying to determine what to set up to make it match. Grr.
Anyway, the weather was really nice today. If only the mud would dry up, we'd be in great shape now.
I'm as far as getting an actual error message now. You have to begin by understanding that in the X world, "server" and "client" are reversed from what you normally think of. So the place where the applications and data reside is called the "client" and your X-terminal is the "server". While it does make a kind of twisted sense, it just confuses every discussion of the subject, and especially so since the writers themselves mix the two ends of the conversation up occasionally.
So the problem appears to be one of "authentication", which is done with what are called "MIT-MAGIC-COOKIES". I've finally learned how this works under circumstances where the user is already logged in at the server end. The authentication cookies are stored in a particular file in the user directory. But when you are using XDMCP to start a connection, no one is logged in yet. So where do the cookies come from and where are they stored? The documentation is of course vague on this, mostly handwaving. But if the two ends of the connection don't agree on the same cookie, they refuse to talk. When one end is Linux and the other is OpenVMS, it gets even messier trying to determine what to set up to make it match. Grr.
Anyway, the weather was really nice today. If only the mud would dry up, we'd be in great shape now.
no subject
Date: 2009-03-17 03:46 am (UTC)no subject
Date: 2009-03-17 10:21 am (UTC)no subject
Date: 2009-03-17 04:31 am (UTC)The really twisted part is that, if you set up a "thin client" desktop, all your clients run on the server and connect to the server on the (thin) client. Confused yet? :P
no subject
Date: 2009-03-17 10:36 am (UTC)no subject
Date: 2009-03-17 04:57 am (UTC)no subject
Date: 2009-03-17 10:40 am (UTC)When Stephen Hoffman goes, the whole DECWorld will go with him. I still like the Alpha and even the VAX. It's a shame that the maneuverings of merely clever market manipulators so often can stifle genuine innovation.