"A User's View of the Workstation" Document/Minicourse Project Questions on the System Model: Fall 1986 Heather Patrick, Jeff Jimenez, Beth Anderson, Bill Bryant, and I have met several times to outline the content of a minicourse/document that would explain to people how to use an Athena workstation this coming September. In addition to the living groups, some 600 students in 2.10, 1.00, and various courses scheduled into 4-035 will be using Athena workstations. We have got a lot of questions. Since everybody else does, too, and since it seems that there is a whirl of activity afoot right now to define the workstation model for Fall, we put together this list for discussion. Our prime goal is to generate a scenario for Fall that a) is feasible, b) has reasonable fallback scenarios, c) answers some of the more pathologic cases that emerge, particularly when asking "How do I go between RT's and VAX's?" and d) is the same for both IBM and DEC hardware. We do not have the resources to document one scenario for VAX's that includes user lockers, and a different one for RT's that does not. Summary The main questions seem to be: 1. Are we using a "straight workstation," that is, a machine without /usr RVD, user lockers, or kerberos? 2. Will the workstation have /usr on RVD? 3. Will Kerberos authenticate logins? 4. Will there be user RVD lockers? Perhaps just course RVD lockers? 5. Will the login process get from Kerberos the name of your user locker and spin it up for you or will you have to do this yourself? 6. How will students spinup course lockers? 7. How much of this is going to have to be pre-doctored into the /etc/athena/rvdtab file on a cluster-by-cluster, course-by-course basis, vs. how much will the student have to supply as up parameters, vs. how much is kerberos going to supply? 8. Will .rvdtab be implemented? 9. How will files be shared between DEC and IBM hardware--particularly at the end of term? 10. How are we going to do backup? Most other questions are detail questions of these major issues. The Detailed Questions Before anyone complains about excessive detail here, understand that we have already decided to tell people the simple version of the story--where everything goes right--first, and only then start to discuss what can go wrong. 1. Are we using a "straight workstation?" This is included because it is a possibility on the RTs. As I said, we don't have the resources to document to widely different environments as would occur if RT's were "straight" and VAX's fully-athenaized. 2. Will the workstation have just /usr on RVD? Again, this is a possibility for the RT's. How, then, will we handle accounts? user real storage space? cleaning out old stuff? users copying files to removable media? student access to course software? 3. Login What will the person see when they walk up to the screen? What wrong states could exist (like no /usr server), and how to correct? When they "press any key," will they see the console login we see now? What's actually going on? Short reboot? Long reboot? What residual situations not fixed by short reboot? How long is this going to take? Will the login window look the same as on timesharing? 3a. Will Kerberos authenticate logins? Where will the logins come from? Userreg? Will they be the same as the "regular" athena login? What about password? What message do you get if wrong username/password? What do you do about it? Can you change your password, does it change in the rest of the non-workstation world AND VV? If kerberos doesn't, what will? 4. Will there be user lockers? This is the big question. No: If there aren't, what are we going to do instead? The questions look alot like 2 above. User lockers can actually be quite independent of anything else. Login could, for example, clean the tree of old user files and mount the user's floppy in the right place, and kerberos could authenticate the user, all without user lockers. Course software is more difficult. Yes: How big? What name? What passwords? Where? And how do we relay this to students? What format (DEC/IBM) by default and how do we know? We have heard it said that we are not going to give people lockers just for courses. We cannot follow this line of reasoning. What if I have a DEC locker and my course uses IBM hardware? We think we have to give people a locker for their course. (And how did the courses get their lockers and check out their software before the term started?) Will $HOME == "~" work? Where will it point? How will users change locker passwords? Will down continue to spindown all lockers on a workstation, even though somebody else may be logged in? What are the mount points for other lockers? 5. Will Kerberos tell the login process the name of your user locker and spin it up for you? Yes: Good. How did it know what to do? What if it wasn't successful? (Server down, some other process has it mounted, how force off?) Where was it mounted? If it wasn't mounted, do you still have a local writeable home directory you can write into? Do you have a local writeable directory even if it WAS mounted? (This is kind of a global "how much is local part of the normal workstation environment" question.) No: Then how do you get it? Do you have to give complete parameters because it's not in /etc/athena/rvdtab. At what point does your personal .login get executed then? If one of these "generic users with generic lockers" approaches gets suggested, then how do we handle security since people have homework and problem set answers around. 6. How will students spinup course lockers? How automatic. Kerberos? In /etc/athena/rvdtab? Fully-qualified up command? Recommend in .rvdtab with up command in .login? How should student cope if nothing comes up? Is there only one course locker, or two in case of trouble? What about all these usual course issues like "how to handle turn-in?" The entire question of how students access course software needs some storylines. 7. How much of this is going to have to be pre-doctored into the /etc/athena/rvdtab file on a cluster-by-cluster basis? This is the "how automatic" question again. I take it there's some idea that there'd be a central database that would somehow know what courses a student was in and get the right lockers spun up, and would be TA-modifiable the way groupadmin works now. Is this what's called the "Service Management System" that Ostlund's working on? What plans are there for this, and which will be available in September. 8. Will .rvdtab be implemented? This is what I think is true NOW: a)anybody can use the up command to spinup a locker anywhere in the net provided that they know all its data, most particularly what server it's on. b)/etc/athena/rvdtab only exists as a list of "default" things to spinup when one issues an up command. c)the statement, "If it's not in /etc/athena/rvdtab, you can't spin it up" is untrue because of a) above. .rvdtab is just like /etc/athena/rvdtab except that when you say up it is referred to first. It has the same format, everything. 9. How will files be shared between DEC and IBM hardware? At minimum, how will you copy between different format lockers? This problem cannot be avoided. Either the student started out with a DEC locker and now is faced with an IBM hardware course, or he took an IBM course and now the course is over and he wants to use his files on DEC hardware. 9a. How will files be shared between users? 10. How are we going to do backup? There seems to be two views of backup. One is the "I make copies of important files I can't stand to lose" on a piece-by-piece basis, as people typically do on micros with floppies. The second is the wholesale copy of all files en-masse in case of catastrophic loss. This gets crossed with the fact we have four different media: IBM RT floppies, IBM RT streaming tape (do we actually have any of these), Micro VAX floppy, and Micro VAX streaming tape. If we propound wholesale catastrophic backup, how does the student retrieve the one file he/she accidentally deleted without copying over their whole file system? What user-friendly utilities does one use to do backup? To do restore? How much does it cost to buy the media? Is it realistic to expect students to shell out $23 for a streaming tape? How do you format it? Have we ordered the media? Made provisions for its sale? Other Questions Groups. Will there be groups for TA's to share files? Mail. Will kerberos be mediating mail incorporation? If it isn't, is there any point in having POP at all? What will you have to do to use POP? Right now there's all sorts of administrative "see this guy" stuff. Will this be taken care of? How do you read your mail on your W20 timesharing system if you inc'd it into your RVD? I understand that, without the notification server, you never get "You have mail" messages. True? When is notification server anticipated? Logout. What does logout do? In particular, does it FORCE spindown of your locker? How is locker spindown forced in the event of a crash? (This is important, because if you have an old x-mounted floating around, you can't use your own files.) Will we have utilities to find out who has your locker spun up? Will we have utilities that let you force them to let the locker go? What state is the machine in after logout? Server crashes. We had a long discussion on what things look like when a) /usr server crashes, b) your locker crashes. The two main cross cases seem to be: i) you were in gnuemacs, ii) you weren't. There are a bunch of black-box messages that come up that could be better. The worst is "Server crashed, spindown, then spinup" or some such which is completely misleading. The other is the retrying problem which freezes things. Is this going to be fixed? What other kerberos mediated utilities will be there? krlogin? krcp? Will there be both krlogin and rlogin, or just an "rlogin" that's the k version? Which standard end-user software will not be available on the VSII and RT? If people are serious about things like RS/1 servers for the RT, how will such servers actually work? Are we going to restrict students to particular clusters? Why? Will rwho show people on other workstations?