CSuite Workshop Minutes - Wednesday, January 25, 1996
David Murdoch opens with agenda ...
- Time
- Distribution
- Hypernews
- Next week
- Majordomo
Trent presents Time management ...
- table I presented on overhead showing headings of:
- Class (dialup, network, unknown)
- Device (tty[a-f], tty[p-r], --)
- Total (##, ##, ##)
- table II presented on overhead showing headings of:
- Class (dialup, ..., ..., network)
- Threshold (76 100, ## ##, ## ##, 1 10)
- Blackout (0, ##, ##, 20)
- Guarantee (120, ##, ##, 60)
A gawk script shown on overhead. Currently this script reads the output
from another script which only runs every 2 minutes updating the # of
each type of user. The percentage of free lines for the class is
computed. The end result sets environment variables to the shell
"tshell" which does the timeout. Very few mods to "tshell" were actually
required.
- Gerard
- What about lynxprog and custom login screen interaction?
- Trent
- This should be easy to do.
- Ed
- Does this "guarantee" imply a second time limit after the guarantee
is up? That is, an "extension"?
- Trent
- Yes, extension is possible, but not yet implemented.
- Gerard
- How do we handle notification of extension?
- Trent
- Having these warnings appear every hour is a nuisance.
- Gerard
- Of course, they should be user-configurable to turn them off.
- David Murdoch
- We need to be careful that what we design isn't too complex to
explain to the user how it works.
- Trent
- We need to be careful not to build the algorithm in such a way that
might encourage usage patterns to shift and all pile up on a
different peak time again.
- Gerard
- How about an indefinite extension with an "n" hour countdown after
the "x" percent saturation point is reached?
- David Potter
- Won't we run the risk of the system being tied up while all of the
users notified at the time the saturation point is reached hold
onto their lines until they are all kicked off at once.
- Gerard
- So you are saying we should have something like a queue, staggering
the kicking off users with extensions according to when they initially
logged in?
- But maybe we are getting too complex. What about just using the
easiest solution to implement.
- How about forgetting the "unlimited" extension and just using fixed
period extensions with warnings which can be turned off by the user
if desired. (e.g. "You've been granted a one-hour extension").
This is simple enough to implement, simple to understand, and solves
the "staggering" problem.
- Ben
- What about "go extend" to actually request an extension if available?
This also solves staggering.
- David Potter
- This might be useful if there is a way for IP's to "bank" a certain
amount of time to get their job done whenever they need it (whether
during periods of high load or not).
David Benoit talks about distribution ...
We've come up with a variable substitution scheme that involves keeping
a parallel directory structure to the /ccn/info documents, say, on
/ccn/src_html/info/, to keep source documents. These documents contain
directives to M4, like _WEBMASTER. At install time, "make" is run to
produce the final document, substituting the directives with M4 and the
variables stored in /ccn/lib/vars.
- Ben
- Do we have a process for requesting a new variable in the works?
Will we extend this model to allowing IP's to use it for things
such as insertion of standard button bars, copyright notices, and
other such features on each page.
- David Benoit
- Sure, like a .defs directory off of the IP directories.
- Ben
- What about standards? Any merit in following NCSA httpd server-
side include syntax where the concept of substitution is so similar?
Even though this isn't done by the server, but is actually just
a preprocessing step done at install time. Yet there are similarities.
Or how about SGML?
- David Benoit
- There is no support for the variable substitution concept either in
SGML or in NCSA httpd server preprocessor directives.
- Ben
- Just so long as we aren't reinventing the wheel in coming up with a
syntax for the substitution variables, as these will be propogated
through the whole info filesystem.
- David Potter
- Rather than widespread scattering of these variables throughout
our documents, we should focus on only needing configuration for
the "index" pages (e.g. of the Help Desk, the first levels of
pages) and just have constant text in the narratives.
- Matthew
- Yes, I agree with David, even though it is lots of work. We need to
keep the global overview, as we need an organized proposal and set
of standards for when we farm out development work to other sites.
Ben talks about hypernews ...
- Have been waiting for latest version (beta .5?)
- In the mean time, David Robinson and I will be reading the "bugs"
page and trying to keep up-to-date. We will try to subscribe, as
well, to some non-tech page for recreation for the purpose of making
postings and getting familiar with the system.
- Until we have a development platform and a new release, it doesn't
seem worthwhile to do much more than this.
Gerard and Chris Maxwell talk about lqtext on Linux ...
- They have it working: one place where the parameters were reversed
from what was expected, which Chris patched, so it works now.
- Will coordinate testing on the CSuite development machine
- David Benoit reminds that it would be good to send back a patch
kit to the lqtext developers for inclusion in the standard distribution.
David Murdoch talks about next week ...
- Gerard volunteers, with help from djm, to do minutes
- David suggests next week we focus on letting other people do
more of the talking who haven't contributed as much
Next week's agenda:
- Gerard to take minutes
- Bulkmail presentation
- Shell programming by Gerard
- CGI programming by David Trueman