|
|
| Author |
Message |
| 25 new of 61 responses total. |
dpc
|
|
response 25 of 61:
| 
Oct 31 20:43 UTC 1997 |
So I assume nobody has any answers to the questions I asked in #21
above? How about some rough approximations?
|
senna
|
|
response 26 of 61:
|
Nov 4 06:51 UTC 1997 |
I always thought the debate was way to overstated considering what was being
argued. I've always thought that the whole thing was kind of irrelevant.
|
janc
|
|
response 27 of 61:
|
Nov 4 12:07 UTC 1997 |
Re #25: Yes, it would be worthwhile to collect usage statistics on
these kinds of things. We do have much of the raw data needed to get
some kind of answer to those questions, though it is buried in huge
masses of other data. I'd be interested in doing that, but it is much
lower priority than getting the 4/670 up or getting the next release of
Backtalk out. I think Mark Conger had expressed some interest in
developing programs for some statistic-generation tasks, but I haven't
heard of he'd found time for that.
|
aruba
|
|
response 28 of 61:
|
Nov 4 23:00 UTC 1997 |
No, I haven't been able to sit long enough to work on those yet. Soon, I
hope.
|
dpc
|
|
response 29 of 61:
|
Nov 8 21:35 UTC 1997 |
Well, janc, can you tell just from glancing thru the data if just
a handful of folks are using Backtalk, or a whole bunch?
|
janc
|
|
response 30 of 61:
|
Nov 9 03:32 UTC 1997 |
The data is not in a form that you can "just glance at". With a little
grepping I can figure out that there have been 208 "hits" on backtalk in
the last week. After some cutting and sorting, I was able to figure out
that these hits came from 48 different IP addresses. I would guess that
this is on the order of 35 to 40 different users.
|
dpc
|
|
response 31 of 61:
|
Nov 11 00:01 UTC 1997 |
Ok, thanx! So we're talking about a quite small percentage of Grexers
coming in thru Backtalk.
|
arthurp
|
|
response 32 of 61:
|
Nov 11 01:00 UTC 1997 |
What percentage of conferencers is that? Careful how you downplay it.
|
dang
|
|
response 33 of 61:
|
Nov 11 03:05 UTC 1997 |
Of course, there's no way to know how many different people ran picospan in
a given month, so we'll never know. :)
|
orinoco
|
|
response 34 of 61:
|
Nov 14 00:26 UTC 1997 |
Well, it seems to me that part of the problem is people aren't yet using
backtalk to it's full capacity. For instance, when and if I get off my ass
and get my web page made, I'm planning on putting in a link to the anonymous
version of the Amalgam conference. I'd be curious to see what sort of use
that'd get...
|
dpc
|
|
response 35 of 61:
|
Nov 14 02:57 UTC 1997 |
I don't think it's a "problem" that very few Grexers are using Backtalk
yet. It's still under development and will take a while to become
popular. The speed is still a major issue. Yesterday at M-Net's
Bod meeting, Dave Thaler, the author of a similar system called
WebYAPP which M-Net uses, discussed speed. Apparently the big slowdown
is when someone starts up WebYAPP it has to invoke a *lot* of M-Net
system processes, so for the present at least telnet is a lot faster
than access from the Web.
|
mta
|
|
response 36 of 61:
|
Nov 14 03:13 UTC 1997 |
That hasn't particularly been my experience with backtalk.
|
orinoco
|
|
response 37 of 61:
|
Nov 15 05:10 UTC 1997 |
Well, backtalk is certainly slower than either telnet or dialin for me, but
perhaps I'm alone in that.
|
valerie
|
|
response 38 of 61:
|
Nov 15 06:48 UTC 1997 |
This response has been erased.
|
orinoco
|
|
response 39 of 61:
|
Nov 15 22:45 UTC 1997 |
That just might do it. :)
|
srw
|
|
response 40 of 61:
|
Nov 19 03:57 UTC 1997 |
I think ISDN has helped Backtalk be easier to use. We know it has some
performance issues, but they are not in backtalk itself, but rather they are
inherent in the stateless nature of an http connection.
Still there are ways to cut down on this overhead, and Jan recently outlined
some improvements he plans in this area. For conferncing a lot, picospan
wins, but for incidental conferencing, backtalk wins because you bypass the
queue and don't have to log in and start picospan.
I have no idea how webyapp compares to backtalk in the area of performance.
|
janc
|
|
response 41 of 61:
|
Nov 19 17:31 UTC 1997 |
I don't know enough about webyapp to say for sure, but I would expect Backtalk
to be pretty similar in speed, perhaps a smidgeon faster. I think both are
stateless (unlike webcaucus), both interpret scripts (though I think
Backtalk's script language is about as fast as it could be, so it might
slightly outperform webyapp there), and both are based on the somewhat
inefficient Picospan file structure (though Backtalk uses some additional
index files to speed things up). My own assessment is that Backtalk still
has some ways to go before it is the best conferencing system around, but it
is already pretty competitive with webyapp. It's probably however that M-Net
has a slightly less overburdened machine right now, which may be more than
enough to consume any basic difference in software performance.
|
orinoco
|
|
response 42 of 61:
|
Nov 19 20:05 UTC 1997 |
'stateless?' What is?
|
valerie
|
|
response 43 of 61:
|
Nov 20 06:04 UTC 1997 |
This response has been erased.
|
mdw
|
|
response 44 of 61:
|
Nov 20 10:06 UTC 1997 |
One of the reasons that PicoSpan has a fairly unsophisicated database,
is that when running, PicoSpan can keep a lot of state around and do
things fairly fast. For instance, whenever you read an item, the first
thing it does is scan through the item to build up a structure in memory
that says where all the responses start for the item.
|
remmers
|
|
response 45 of 61:
|
Nov 20 10:23 UTC 1997 |
And then the in-memory structure persists as long as the user
"stays in" the item, so that going back and re-displaying
previous responses is fast. A web-based conferencing program
can't keep stuff in memory between accesses because the web
just isn't designed that way -- any host resources allocated
to a user go away as soon as a page download is finished.
|
dpc
|
|
response 46 of 61:
|
Nov 22 18:07 UTC 1997 |
Let's assume there's no queue. Does the discussion about web-based
conferencing being "stateless" mean that web-based conferencing
will always be slower (and hence less satisfactory) than telnet-
based or dialin-based conferencing?
|
scott
|
|
response 47 of 61:
|
Nov 22 19:54 UTC 1997 |
Depends on your definition of "satisfactory". Web-based systems let me type
at speeds limited only by my own PC speed, use client software that I like,
and use client software that is friendlier to (for instance) voice dictation
software than a term program would likely be.
|
janc
|
|
response 48 of 61:
|
Nov 23 05:22 UTC 1997 |
It should be possible to build web-based systems that are just as fast as
text-based systems. Faster even.
|
mdw
|
|
response 49 of 61:
|
Nov 23 06:38 UTC 1997 |
It may be quite difficult to design something as efficient (frugal with
system resources) using a web interface as with text. Designing something
that is "just as fast/faster" is not a problem if you assume hardware
that is sufficiently cheap & fast. There are some messy GUI
issues, however, that make this a bit less than straightforward.
|