|
Grex > Coop11 > #145: The Automatic Idle Keyboard Disconnect | |
|
| Author |
Message |
prp
|
|
The Automatic Idle Keyboard Disconnect
| 
Dec 11 19:19 UTC 1999 |
Grex currently has a timer that automatically ends a connection when
there has been no keyboard input for fifteen minutes. I find this
limit too short for several reasons, among them:
- It takes longer than that to read what "man whatever" prints,
at least if you try and understand it.
- It can take longer than that for a phone call, and a couple
of stops on the way back.
So, I propose that the limit be raised to one hour. Computers
should be patient with people. Various programs that have time-outs
measured in seconds drive me crazy, but that is another item.
Some will object to this on the grounds that it will lengthen the
queue waiting for access to Grex. This is not a problem. The
number of network connections can be increased by the expected number
of connections idle from sixteen minutes to an hour. The combined
result would be no change in the queue length. I would estimate the
expected value at one to start.
I seem to recall seeing something about this somewhere else, but
I looked for it and couldn't find it. Sorry.
In summary:
- Increase the idle keyboard time-out form fifteen minutes
to one hour.
- Increase the maximum network number of network connections
by one.
|
| 58 responses total. |
pfv
|
|
response 1 of 58:
|
Dec 11 19:24 UTC 1999 |
This response has been erased.
|
prp
|
|
response 2 of 58:
|
Dec 11 19:26 UTC 1999 |
Why?
|
don
|
|
response 3 of 58:
|
Dec 11 19:49 UTC 1999 |
I have a better idea: modify idled such that when there is no queue, let
people be idle for an hour. This fixes the problem.
|
prp
|
|
response 4 of 58:
|
Dec 11 19:56 UTC 1999 |
Not if there IS a queue.
|
don
|
|
response 5 of 58:
|
Dec 11 20:19 UTC 1999 |
If there is a queue, then extending to 1 hour would increase the wait; fixing
this by adding more connections would use up more cpu time and memory. So
extending when there isn't a queue should be the best option.
|
pfv
|
|
response 6 of 58:
|
Dec 11 20:23 UTC 1999 |
This response has been erased.
|
prp
|
|
response 7 of 58:
|
Dec 11 20:40 UTC 1999 |
Re 5: No, adding both an additional connection and stopping an
existing one doesn't use up any more cpu time. It does take
some memory, but so little that it doesn't matter.
|
pfv
|
|
response 8 of 58:
|
Dec 11 20:56 UTC 1999 |
If ANYTHING, I'd like the "warning" to start a noisy countdown.
One blurb to the screen, what..? 5 minutes before you die (15?)
is sometimes insufficient to grab your attention. (like debriefing
the wife in the next room ;-)
Be nice too, if we could do something _drastic_ about multiple
ports.. For a short period, like 15 minutes or so, they can be
REAL HANDY.. But grex has folks that use them NORMALLY - and it's
_supposedly_ "frowned upon".
|
sno
|
|
response 9 of 58:
|
Dec 11 21:47 UTC 1999 |
Re-logging in is not that big a hassle, even with a que. Right now
the time is twenty minutes, fifteen minutes to the first warning and
five more to get done in the bathroom.
With Grex now having a que active most of the time now, extending the time
would be unfair.
|
gull
|
|
response 10 of 58:
|
Dec 12 01:14 UTC 1999 |
Re #5: Actually, it would't increase CPU use; idle connections are, almost
by definition, consuming practically no CPU, unless they're running a
compile or something. You'd be essentially trading idle connections for
active ones, if the balance works like the original poster thinks it might.
Personally, I don't generally have a problem with the idle timer; it doesn't
take me 15 minutes to read one screenful of text, though.
|
gelinas
|
|
response 11 of 58:
|
Dec 12 02:14 UTC 1999 |
I dunno. I can see taking more than fifteen minutes to get through a
manpage, or even an item, but less requires keypresses every 24 lines or so
(on my screen, anyway), so it's not much of a problem. On the other hand,
the warning notice itself can be inconenient.
This isn't my main (or only) Un*x system, so my having to wait to get
back in is not that annoying. Others may have different needs.
I've also set my dial-up software to hang up after idle 30 minutes.
|
prp
|
|
response 12 of 58:
|
Dec 12 16:54 UTC 1999 |
Re 10: are there any Man entries that produce only one page of text?
Re 9 and others:
This will NOT increase the length of the queue.
I know queuing theory is sometimes counterintuitive, and "expected
value" is from math class, but read carefully and you'll see that
this really is true.
|
pfv
|
|
response 13 of 58:
|
Dec 12 17:10 UTC 1999 |
It has nothing to do with queue-length, it has to do with wasting
a port, I suspect..
|
gull
|
|
response 14 of 58:
|
Dec 12 18:10 UTC 1999 |
Re #12: No, most man entries are more than one page. But it takes a
keystroke to tell the pager to go from one page to the next, so unless it
takes you more than 15 minutes to read each page, you won't time out.
|
scg
|
|
response 15 of 58:
|
Dec 12 18:55 UTC 1999 |
Grex's port limits are an artificial construct. We could allow pretty much
any number of ptys, at least up to 256 or 1024 or something like that, pretty
easily. The limit on the number of ports in use is based on an estimation
of how much of Grex's other resources a typical user consumes. If users were
to suddenly start using a lot less resources, we could easily let more users
on at once.
|
other
|
|
response 16 of 58:
|
Dec 12 22:56 UTC 1999 |
just pipe man pages through more, and you won't have a problem. :):)
|
prp
|
|
response 17 of 58:
|
Dec 13 19:01 UTC 1999 |
Re 14: Gull assumes I use more or less as a pager. I don't. I
use my terminal emulator program. This may be why I run into
this annoyance more than most people.
|
pfv
|
|
response 18 of 58:
|
Dec 13 19:18 UTC 1999 |
Dialin? doze-telnet?
Whatever it is, I believe yer all of the 2nd person to complain -
and for essentially the same reasons. Most folks know to stay busy
or idle-out. (I suspect more than a few defeat the idler, though).
|
gypsi
|
|
response 19 of 58:
|
Dec 15 22:54 UTC 1999 |
I hear the beep for 15 minutes and just hit "enter" since I'm usually idling
at a bbs prompt.
If you don't want to or can't use "more" as a pager, then oh well. Not many
people have complained about the twenty minute idle-out time...you're actually
the first or second I've seen.
If people idled for an hour, nobody would ever get on. I remember the old
days before the queue...oi vey.
|
carson
|
|
response 20 of 58:
|
Dec 16 00:55 UTC 1999 |
(yep, I was there, and it was horrible. long busy signals, constant
waiting while the comm program redialed... and then discovering that
*all* *six* phone lines were filled with people in that damn "party"
program.)
(ok, maybe not those old days.) ;)
|
gracel
|
|
response 21 of 58:
|
Dec 21 03:05 UTC 1999 |
There have been occasions when I wished the time-out thing were just a little
longer, but it's due to bad time management on my part (half-way through
composing a complex e-mail message, realize the time & run off to be a
crossing guard for 10 minutes -- return to the computer just as it logs me
out). On the whole, the present time limits seem reasonable, and an
hour is a long long time.
|
prp
|
|
response 22 of 58:
|
Dec 21 21:55 UTC 1999 |
Re 19: I hate to keep saying this, but people seem to keep missing it.
The proposal WILL NOT INCREASE the queue.
|
eeyore
|
|
response 23 of 58:
|
Dec 22 08:58 UTC 1999 |
I think that the major thing at this point is that there are few people that
see the need for it at all in the first place. 15 minutes is plenty of time
to go to the bathroom, throw a pizza in the microwave, or grab that cd that
you really need to listen to while Grexing.
|
cmcgee
|
|
response 24 of 58:
|
Dec 22 15:29 UTC 1999 |
I'm with eeyore. I see no reason, if there are ANY Grexers waiting in a
queue, to allow other Grexers to extend the amount of IDLE time, just to keep
from having to get in the queue again. If you need more than 15 minutes, let
someone else on, and get back in the queue. (actually, as I recall, you
actually get 20 minutes).
|