chuq@plaid.Sun.COM (Chuq Von Rospach) (03/02/88)
>I am going to urge a NO vote on comp.binaries.hypercard. My reasons? >Basically, two. First, the volume would be outrageously large - stacks >aren't small. This is a value judgement. I'd rather the net create the group and let individual sites decide whether it's worth carrying. Considering the number of Mac users on the net, it seems like a worthwhile service. >Second, the limited applicability of this to the needs >of the net as a whole. While there are a lot of Mac owners on Usenet, >there are also many, many more who do not own Macs, or do not own Macs >on which Hypercard is usable. Hmm. According to Brian's stats, 37,000 folks read the largest newsgroup, news.announce.conference. if we use that value as 1 (the ideal largest practical population, we see some interesting comparisons). For instance: comp.unix.wizards, at 21000, rates a .56 comp.sys.mac, at 21000, also rates a .56 comp.sources.games, at 18000, rates a .48 comp.sources.unix, at 16000, rates a .43 comp.binaries.mac, at 10000, rates a .27 comp.sys.mac.hypercard, at 8000, rates a .21 What's this mean? I'll let folks come up with their own general conclusions, but I find the following things interesting: o as many people read the network for Macintosh stuff as for Unix stuff. This implies (and I've said this before, and I'll say it again) that this is no longer a Unix Network, but a Network hosted on Unix machines. A very important semantical difference that people need to keep in mind when they look at the future of the net. It isn't Unix First anymore. o This is despite the fact that comp.sys.mac has a heavier traffic level. It is generally considered as a truism (and we can argue this if you want, I'm easy) that the more messages and volume in a group, the more folks who will unsubscribe. volume in comp.sys.mac is something like 5X the size of comp.unix.wizards. This implies that (1) mac folks are a lot more motivated about their material, so they don't unsubscribe as soon, or (2) the Mac population is a LOT larger, potentially, but held down because of the size of comp.sys.mac. We'll see what the splittings that are happening in that area do -- I'd be unsurprised to see the Mac groups have a significantly higher readership than the unix groups in six months. o everyone wants to play games. o Half the folks who read unix-wizards bother to look at Unix source. o only one-third of the Mac folks bother with Mac binaries. o 40% of the folks who read the Macintosh group also read the hypercard group (implying they either have or are interested in getting hypercard). That's not a trivial number, especially when you remember that HC is only six months old.... o Depending on how you skew the numbers, you could claim that (1) a more focused binary group (comp.binaries.hypercard) would let more people rejoin comp.binaries.mac AND there would be a higher percentage of comp.sys.mac.hypercard readers reading comp.binaries.hypercard; or (2) you could claim that because HC binaries are large, fewer people would hassle with them. I personally have no idea which one is right or wrong, so you can choose the one that fits your argument best. But taking comp.sources.unix as a highwater (50%) and comp.binaries.mac as a low-water (33%) we can imply that comp.binaries.hypercard would be of immediate benefit to between 2500 and 4000 people. This excludes any benefit that might accrue to the other binary group by getting the hypercard stuff out of it (which was, in many ways, one of the primary reasons for starting this in the first place). So, 2500 to 4000 people. Is it worth it? I think it is necessary IF we want the comp.binaries.{mac,hypercard} stuff to work right. Strictly speaking, there isn't enough room in that one newsgroup for both sets of files. If we don't feel that this is the proper way to go, then we should consider getting rid of binaries completely. We should support them properly, or not at all. Side thought: interesting numbers. If you compare the readers of the group vs. the readers of the binary group, you get some very interesting numbers. For Mac's, it's about 35%. For amiga, 40%. For IBM-PC machines, 60%. Which means that, at best, a binary posted to the net will reach half the folks who are interested enough to read the main group. That doesn't say anything about the number that actually download it. It's impossible to get good evidence on that, but I have one case of empirical data I'll pass along: After one posting I made (it was about five parts) I was inundated by a dozen requests for parts. I finally put out a call saying "Did ANYONE get it whole?" -- and the answer was a definitive yes, the famous "I didn't get part 3" messages are a small subset of the network (fortunately). But of that, I only got about 50 notes from folks who had downloaded it and were able to use it. Now, even assuming that back then things were half as large as they are now, and not everyone bothered to reply, we can fudge the numbers all day and still get maybe 200 of 4000 people using that program. Now, that would translate to 400 of 8000, say. I'd say, thinking about that (I never put those numbers together this way before....) that you're lucking if you can get 25% of the readers of a group to download a given program. A better number might be 10%. On something like Delphi or CompuServe, where there are long-term archives (something USENET needs desperately and doesn't have, with no practical way of adding them) and you're still lucky if you can get 10% of the users to download [albeit at a fee -- USENET is 'free' except to the folks who really pay the bills]. so, 40% of the readers of a support group read the binary group, and 15% download a given program [yes, I'm rounding numbers all over the place. Come up with better numbers yourself!] So, out of 21000 comp.sys.mac readers, a given binary posting, large as it is, is useful to about 1500 people. So I'm asking: Is it worth it? Do binaries belong on USENET? This isn't a rhetorical question, this is the first time I've frankly crunched out the numbers, and as strong a supporter of these things as I've been, I'm fascinated by what I'm seeing. And I'm not sure than binaries are justifiable based on the size/useage tradeoffs. Also, we have to remember that with VERY few exceptions, this stuff is all available elsewhere, too, for the cost of a modem and the time it takes to access a BBS, or CompuSserve, or GENie, or Delphi, or whatever. Or by going to user groups. How many of that 1500 people have no other access? Or would be unable (or unwilling?) to get access to another distribution point if this one was cut off? I dunno. Alternatives: o Do nothing. continue to allow binaries. Yes, they're worth it. o Ban Binaries. Nuke the groups. No, they're not worth it. o Archive Servers. Instead of moderators, find people to put up binary archive servers. Post indexes and notes about new programs, let folks who want them have them sent directly, as an automated mailing list type of thing. If usage/interest is really that low, this might drop netwide traffic (and shift traffic to specific high-usage links) o something else. Got any good ideas? I want to run a straw poll. binaries, yes or no, with whatever feedback you might want to add. I've created binary-poll@plaid.sun.com for this. (also sun!plaid!binary-poll, or whatever variant works for you). Send me a yes/no on binaries, just so I can see what people think about it. If you've got feedback you want to add, pass it along, or post it to news.groups so we can sit and chew on this a bit. Interesting. Isn't playing with numbers fun? Chuq Von Rospach chuq@sun.COM Delphi: CHUQ There is no reason for any individual to have a computer in their home. Ken Olson, President, Digital Equipment, 1977
longstaf@lll-lcc.aRpA (Thomas A Longstaff) (03/03/88)
Very interesting post. Given only the three choices of do nothing,
nuke the groups, or archive servers, I would vote for do nothing.
I am actually in the situation where I have access to a "real" network
(as do you, given the address of this mail), so I can easily get to
archive services. For this situation, simply pointers to new
additions to the archive are sufficient (when the net is going well,
but thats a different story...)
However, there are at least a reasonable number of *nodes* (not users)
that do not have this kind of access. If binaries were cut off from
these nodes, the ease of getting and trying out the binaries
(including hypercard stacks as a special case) would drop drasticly.
I'm not saying it couldn't be done, but it would take additional
effort and for many of the posts, the usefulness is not obviously
worth the effort.
Perhaps the best way to keep everyone happy would be to post summaries
of binaries and their archive location to nodes on arpa-type nets, and
continue to post binaries (at the nodes discretion of course) to
other, less connected, nodes. Alternatively, if many nodes are in the
former category, some sort of automatic downloading of requested
binaries from the archive could be set up such that the first user at
a node requesting a binary would cause a copy to be stored (at least
for a short time) on the node. Others would get this copy rather than
cause another download. In some form, this could even work for the
less connected nodes.
So much for my 2 cents.
Tom Longstaff
longstaf@lll-lcc.llnl.gov
{lll-crg,harvard,sun,dual,rutgers,seismo,ihnp4}!lll-lcc!longstaf
Lawrence Livermore Natl Lab
L-542, Box 808
Livermore, CA 94550
415-423-4416