1. Tony Scott agreed that our proposed method of working - one user running a standard program such as VIEW while another user edits, compiles and runs programs - was acceptable. He emphasised that the system was really geared to one editing user and a number of separate real-time jobs. Any attempt to enable two users to develop tasks simultaneously would have to be done by ACL.
2. A number of problems so far encountered were mentioned. These are listed below, together with explanations and actions.
3. INSTALL - currently, if a new task is installed, the disc directory is not updated. If the system crashes, the new task, and sections of the disc, are lost. To avoid this, the system must be SAVED, but this is only possible if no other user is working.
Tony replied that the problem arises because SAVE copies core direct, and no I/0 must be pending. Alan Reid (UKAEA) had modified RSX PLUS so that INSTALL automatically did a mini-save which did not require a quiescent system. Tony would assist in getting the mods into RSX PLUS III.
4. JRG asked if it was possible to provide INSTALL from both consoles. It appears this is OK, since both will use the same core (must run in lower 32K) so there cannot be any conflict.
WDS suggested that the loss of disc space could be overcome by giving a name to a saved area. Tony thought this was a good idea.
5. UIC - Under RSX, all files are created in the same user area (RSX), and there is no LOGIN facility. It is possible to access other areas by changing pointers in core (!). Tony said that the problem stems from having only one LUN table for all users. It might, however, be possible to provide LOGIN as a task under both TDV and MCR.
6. Error Messages - All error messages from FORTRAN, and system errors, are printed on the MCR console, even though they may have been initiated from TDV. This was again caused by having only one LUN table. The FORTRAN messages could be re-routed by providing two FORTRAN libraries. It might be possible to provide a pseudo-handler on the error LUN slot, but most of the information required has been lost at this point. It was also noted that MCR can lock out all messages.
7. Reassigning LUN slots - Currently, only MCR can reassign LUN slots, although TDV does require this function. Again, the problem is caused by having only one LUN table. Tony said that it would be possible to provide the function under TDV, but that care would be needed to prevent clashes. A possible solution would be to limit TDV to use certain slots only, and this would require changes to programs developed under TDV, but installed under MCR.
8. VT ON and OFF - This facility is missing from RSX and RSX graphics, Tony agreed to find out why.
9. DEFINE - the DEFINE routine to create a random access file does not allow file extension. Tony explained that, whereas DOS DEFINE produced a linked block file, RSX DEFINE produces a contiguous area. He agreed to find out what happened if DOS tries to read or delete such a file.
10. Clear I/O - real-time programs often allow input from a number of different sources, and, when they receive input from one, wish to clear the other requests. RSX queues al1 I/O, and has no priority queue, so it is unable to cancel a request. Tony and WDS are discussing this.
11. FORTRAN library- the bugs in this will be cured when RSX PLUS III is released.
12. FORTRAN calling conventions - it appears that changes have been made to the way FORTRAN passes array arguments to subroutines. Tony did not know of any change, but agreed to look into it.
13. Tony noted all the points raised and agreed to inform DEC of our requirements, although he was not very hopeful that Maynard would change anything.
14. He suggested that he could assist us in modifying sections of RSX (such as INSTALL, LOGIN, re-assigns, etc) after the RSX course This help would come out of the 5 days consultancy that ACL has purchased with RSX PLUS III. No formal assistance over and above this has been agreed. He intended to limit the practical aspect of the course, and to go deeper into RSX workings. A knowledge of DOS would be assumed.
15. He commented that, for the PD11, RSX had been altered to solve most of the problems. In particular, there was one LUN table per user and LOGIN facility. This did mean that RSX11 required twice the core that RSX15 uses. He agreed to send ACL a RSX11D manual for comparison.
