J.Crowcroft@ucl-cs.UUCP (11/15/89)
From: Jon Crowcroft <J.Crowcroft@uk.ac.ucl.cs> Using the ISODE Transport Service code, i have succesfully built an X server and a few clients! These a strictly experimental (like an existence proof) and not very heavily tested at all, but a couple of interesting things came to light in the process: 1. putting transport in does not appear to impair the performance noticably. 2. It does increase the process size by O(100k) per process, so if you're short on memory, you might want a smaller transport (or else wait til it appears in U**x kernels...). 3. The X interface to TCP/DECNET was clearly well written, since it fell out quite nicely onto ISODE transport service interface. 4. Security and Nameing in X need some overhaul - the :/:: distinguishing of TCP/DECNET is unsatisfactory...the per host access control is too simple. the XHost structure aint big enough for ISO addresseses!! any questions? cheers jon
rws@EXPO.LCS.MIT.EDU (Bob Scheifler) (11/16/89)
the :/:: distinguishing of TCP/DECNET is unsatisfactory... You are welcome to suggest a syntax for OSI. the per host access control is too simple. Of course, it was intentionally simple. There's a hook for alternate mechanisms, use it (we certainly do). the XHost structure aint big enough for ISO addresseses!! Huh? I don't know what an XHost structure is. The XHostAddress structure in Xlib takes a pointer to an address, I don't see any limitations there. The HOST structure used inside the server explicitly states that it will need to be bigger eventually, and in any case is an internal implementation structure.