|
|
| Author |
Message |
| 7 new of 44 responses total. |
rcurl
|
|
response 38 of 44:
| 
Dec 12 00:38 UTC 1998 |
Would it really? What fraction of all the uids are essentially idle?
Just hogging OS space? If they were swapped out and then swapped back
in at the users time expense when called upon, there could be more uids.
The questions would seem to be balancing expanding the possible number
of uids at the cost of slight speed decrement - and exactly where that
balance occurs.
|
scott
|
|
response 39 of 44:
|
Dec 12 00:40 UTC 1998 |
The programming involved would dwarf the effort to move to an OS with more
UIDs.
|
aruba
|
|
response 40 of 44:
|
Dec 12 03:08 UTC 1998 |
Re #38: You would still have a big problem, Rane, if two users shared the
same UID. They could delete each other's files, and what happens if they're
both on at the same time?
|
rcurl
|
|
response 41 of 44:
|
Dec 12 03:28 UTC 1998 |
Oh...they couldn't be told apart in the OS...down in flames and back to
the drawing board... 8^{
|
scg
|
|
response 42 of 44:
|
Dec 12 03:42 UTC 1998 |
Right. There are a number of things that UIDs are used for. One of those
is that every file on the system has the owner's UID associated with it. For
example, when you do an ls -l and see the list of files with their owners,
ls is actually reading off the disk a list of files and UIDs, and then looking
up the UIDs to see which user is associated with them before displaying the
listing.
|
mdw
|
|
response 43 of 44:
|
Dec 12 04:22 UTC 1998 |
It would be *much* easier to switch OS's than to try to engineer a
32->16 bit UID mapping scheme. It would be possible, and I spent a lot
of time thinking about this for UM, but it's really not worth it for
grex.
|
krj
|
|
response 44 of 44:
|
Dec 12 04:54 UTC 1998 |
Rane's resp:38 leads me back to thoughts of the 1970's Control Data
NOS operating system, as modified by MSU... we do NOT want to get
into operating system customization any more than we need too...
(ah, nostalgia...)
|