rick@ameristar (Rick Spanbauer) (05/25/90)
In article <9005240215.AA01426@psi.com> schoff@PSI.COM ("Martin Lee Schoffstall") writes: >> Thus, I >> applaud the efforts of CERFnet and UUNET to offer authenticated, >> for-fee access to Internet as a whole. I can guarantee that I'll >> be interested in using their services whereas Marty's (if I understand >> him correctly) just looks like an interesting service. > >You don't understand me correctly. Essentially we've been waiting >for the time that a prescedent is set, permission is granted, or >a consensus may be achievable in the Internet community. That time may >be now, and the PSI blitzkreig may descend. The plight of many small technical businesses is that we just cannot justify spending $30K+ for access into the Internet for the occasional FTP/smtp transfer. Were access fees brought inline with the level of service offered, eg $2K-5K/yr for dialup SLIP is reasonable, surely PSI and other regionals would see their business pick up substantially. Note that it isn't small business alone that has a problem with the high connection costs to the Internet. Ameristar sells IP/TCP network products and every once in a while I ask some of our larger customers ($20M & up) why they are not on the Internet. The answer is usually that the perceived value of the connection is not in line with the yearly access fee. In such cases, a low cost dialup SLIP service would go a long way in giving people a chance to experiment with Internet access to evaluate its usefullness to their organization. Dialup SLIP is also a safe way for the regionals to toy with their price/volume curve without having to add infastructure (ie additional or higher capacity links) at the outset of the experiment. One other suggestion I have is that the regionals ought to survey potential customers about the sort of connectivity and services they would purchase as a function of cost. Good starting sample data sets might be the lists of technical companies that local business organizations or government maintains, or even UUCP maps. >I'll of course let you know in the least crass commercial manner that >I can, if it is of interest..... Please do make an announcement of any new services PSI introduces. I for one would like to hear what PSI is doing to moderate access fees without having to ping your sales organization every few months. >Marty Rick Spanbauer Ameristar Technology Usual disclaimers: my opinions are my own, etc.
sl@van-bc.UUCP (Stuart Lynne) (05/26/90)
In article <1990May25.163528.14300@ameristar> rick@ameristar (Rick Spanbauer) writes: >In article <9005240215.AA01426@psi.com> schoff@PSI.COM ("Martin Lee Schoffstall") writes: } The plight of many small technical businesses is that we just cannot } justify spending $30K+ for access into the Internet for the occasional } FTP/smtp transfer. Were access fees brought inline with the level of } service offered, eg $2K-5K/yr for dialup SLIP is reasonable, surely PSI } and other regionals would see their business pick up substantially. van-bc has a dialup SLIP link (9600 bps) into BCNet and it works quite well. It's low cost, but effective for an NNTP news feed, mail and the occasion FTP. We have configured the modems (T2500's in V.32 mode) to auto-dial our private phone number on DTR. Every ten minutes we run a script which tests if the link is alive. If not we turn it off and on again. This drops the line and redials it. Works fairly reliably, and the cost is low. Now that we have it we couldn't do without it. -- Stuart.Lynne@wimsey.bc.ca ubc-cs!van-bc!sl 604-937-7532(voice) 604-939-4768(fax)
schoff@PSI.COM ("Martin Lee Schoffstall") (05/27/90)
Rick, Someday I hope to meet you..... 5 years of private and public email and our paths haven't crossed yet, just swords.... :-) The plight of many small technical businesses is that we just cannot justify spending $30K+ for access into the Internet for the occasional FTP/smtp transfer. Were access fees brought inline with the level of service offered, eg $2K-5K/yr for dialup SLIP is reasonable, surely PSI and other regionals would see their business pick up substantially. Aside for PSI/PSINet not being a regional, (think about th N word), I think most of us have recognized an entry level market. There must be half a dozen providers who have something that looks like ~$10K with dedicated facilities. One of the issues on the dialup front is that a major cost is hidden, the message units that all businesses have to pay. In many LATA's a 4wire unconditioned analog circuit costs $100/mo, the cross over point where the message units are more expensive is probably not too many of the 3 hour periods described below. Somehow dialup Internet access and SMTP don't go hand and hand in my mind, my estimate is that your going to have keep a connection open for about 3 hours every day to have some probablity of synchronizing with all the SMTP agents pushing mail out of their queues for the site. Realistically you'll be running uucp/tcp to a site like UUPSI who is MX'ing for your domain. Note that it isn't small business alone that has a problem with the high connection costs to the Internet. Ameristar sells IP/TCP network products and every once in a while I ask some of our larger customers ($20M & up) why they are not on the Internet. The answer is usually that the perceived value of the connection is not in line with the yearly access fee. In such cases, a low cost dialup SLIP service would go a long way in giving people a chance to experiment with Internet access to evaluate its usefullness to their organization. Dialup SLIP is also a safe way for the regionals to toy with their price/volume curve without having to add infastructure (ie additional or higher capacity links) at the outset of the experiment. At least you could talk them into getting a good quality UUCP connection so they can do email. I'm frightened by the lack of participation of many vertical industries in communicating with their customers, suppliers, except through phones. (Have you ever entered voice mail grid lock, where neither party ever gets through due to synchronization problems, and the use of voice mail systems as filtering devices. Someday the only time you'll ever get an answer is by dialing 1-900.lovenow). One other suggestion I have is that the regionals ought to survey potential customers about the sort of connectivity and services they would purchase as a function of cost. Good starting sample data sets might be the lists of technical companies that local business organizations or government maintains, or even UUCP maps. This has been done, but your sampling focus is a good suggestion. Please do make an announcement of any new services PSI introduces. I for one would like to hear what PSI is doing to moderate access fees without having to ping your sales organization every few months. Maybe Kent England will radio you in reports from "The Front", rumored to have been named "Operation Fortress Beantown", BarHarborAirlines has guaranteed that they can airlift as many cisco's as they will need to sustain the battle. :-) Marty
wbe@bbn.com (Winston Edmond) (05/28/90)
In article <90121@uunet.UU.NET> rick@uunet.UU.NET (Rick Adams) writes: >UUNET plans to offer access to any Internet site from any Compuserve Dialup >in the continental US. >All access will require an individual login/password for accountability and >authorization to use the Internet will be verified before the account is >established. >We plan to charge $5 per connect hour (in increments of 1 minute). Would UUNET be a CompuServe service provider, charging a $5/hr. surcharge, or are you talking about using the CompuServe network and somehow bypassing logging in to CompuServe itself? Is the individual login/password the usual CompuServe user number and password, or a UUNET login? >We could start tomorrow if I could figure out how to bill for it. >Its not worth sending out invoices for only $5. There are a number of CompuServe service providers that collect surcharges directly from CompuServe -- the user gets a single bill, and CompuServe's billing system keeps track of the surcharges in case the user has questions about his bill. This allows a service provider to charge small amounts, like the $.02 stock quote charge, or the $.90 car profile charge, and still be able to make money -- you don't have to bill the customer yourself. -WBE
mckimg@bronze.ucs.indiana.edu (geoffrey mckim) (05/29/90)
>In article <NELSON.90May21230840@image.clarkson.edu>, >nelson@sun.soe.clarkson.edu (Russ Nelson) says: >>That's too bad, because when I've gone travelling in the past, I've called >>the local university, asked for their terminal server's phone number, and >>telnetted back to Clarkson to check my mail. It's a shame that that kind >>of service has to go away... > >Actually, you can still check your mail, but for the price of a toll call >back to your own terminal server. > >Maybe this is the price we have to pay for added security? > >/Pete >-- >Peter M. Weiss I don't mean to flame but... Obviously one of the primary benefits of something like the Internet is fast, efficient connections around the world. Sure, if we wanted to, we could all just have cheapo 1200 baud modems on our desks and dial up whatever machine we want to directly. But that sort of defeats the purpose of a high-speed network. In other words, the easiest way to improve security is to simply disconnect all the machines on the network from all others. But then we've got no network eh? I'm afraid that knee-jerk reactions have long been the hallmark of those in charge of computer security. I realize that it will always be difficult to balance functionality and security, but I also hope that people realize that the reason for the network's existence is FUNCTIONALITY. I for one will certainly work to fight the elimination of dial-up terminal servers connected to the internet. Let's make our hosts more secure and not intentionally cripple the internet. Geoffrey McKim *** Standard disclaimers apply *** Indiana University
J.Crowcroft@CS.UCL.AC.UK (Jon Crowcroft) (05/29/90)
>I think that "fast connections around the world" does not mean that we >have to allow anyone with a modem and terminal to telnet/rlogin into any >host at will on the Internet. Doesn't compute. dont agree - we have just been checking how fast you can clock a line based on the copper used for UK phones, and managed 200kbps without much trouble - i.e. multiple basic rate ISDNs (without even needing much tcp header compression) will make a X and NFS to 20 million UK homes feasible before the year 2000 - if anyone thought building such a wacky service ... if they do, security on my VCR is gonna be ultra-tight (i dont want some hacker making me miss my favourite show)... jon {p.s. this is not entirely serious:-}
kwe@bu-it.bu.edu (Kent England) (05/30/90)
In article <9005270423.AA19852@psi.com>, schoff@PSI.COM ("Martin Lee Schoffstall") writes: ... > At least you could talk them into getting a good quality UUCP connection > so they can do email. I'm frightened by the lack of participation of > many vertical industries in communicating with their customers, suppliers, > except through phones. (Have you ever entered voice mail grid lock, where > neither party ever gets through due to synchronization problems, and > the use of voice mail systems as filtering devices. Someday the only > time you'll ever get an answer is by dialing 1-900.lovenow). > ... > Please do make an announcement of any new services PSI introduces. I > for one would like to hear what PSI is doing to moderate access fees > without having to ping your sales organization every few months. > > Maybe Kent England will radio you in reports from "The Front", rumored > to have been named "Operation Fortress Beantown", BarHarborAirlines has > guaranteed that they can airlift as many cisco's as they will need to > sustain the battle. :-) > > Marty Glad to! [BTW, it's not "Operation Fortress Beantown". Only folks from elsewhere and local sportswriters call Boston Beantown. :-] We got the three inches of water pumped out of the main bunker and the dehumidifier is fixed now, so it's bearable down here. I have a pretty good view of the harbor and Logan airport. PSI has been airlifting in dial-up servers, but we shot down most of them, since the visibility was pretty poor and PSI didn't pay for good IFR gear. Better luck next time, Marty. :-) Seriously, though, I share Marty's concern about a lack of interest in commercial enterprises in supporting business using modern internetworks. It would seem obvious to me what the benefits are, so perhaps we just need to be more aggressive in touting the capabilities of the technology we know and love. Marty will see to that, as far as he can, since his livelihood depends on it. I see a definite interest on the part of commercial organizations wishing to join the research and education (R&E) Internet, but I see much less interest from them in joining a purely commercial internet service- they don't seem to understand who they would be able to talk to on a purely commercial internet. Come to think of it, neither do I. But certainly there is a lot of interest from commercial outfits in becoming a part of our R&E Internet and in exploiting that technology to talk to the existing base of constituents. NEARnet has had a phenomenal response from commercial organizations joining NEARnet and the Internet. There are a lot of research labs in the Boston area, and they are very eager to be a part of the R&E Internet. Perhaps it is just that they are able to move more quickly allocating budget to the NEARnet connection, and are therefore able to connect more quickly than some of our colleagues in the academic field in New England. We'll see. Like the PSI folk, NEARnet is aware of an untapped potential base for lower cost access to the Internet, from both the commercial organizations and from academic organizations. As might be expected, this is due mostly to a scale pyramid= there are simply more small sites than large. NEARnet is committed to broadening its base of support to include smaller sites at lower cost. We are chafing against technical issues in being able to offer high quality service at lower costs, since we balance hardware and people costs, and people costs are not really that bandwidth sensitive. We also have strict standards on quality of service, and we do not wish to compromise these standards in offering less costly access, since we know that later on our customers would regret the compromise as much as we. Then there is the issue of just what is part-time access; is it terminal dial-ups, SLIP/PPP (host or router?), uucp, what? We have to keep in mind what services we provide and we have to make sure that our clients understand what they can and can't do with new service offerings. All in all, though we have been able to keep costs quite reasonable, so far as we can tell in comparing costs with our neighbors. As far as keeping up with new service offerings, you should keep an eye on comp.newprod. PSI always posts announcements of new service offerings there, and NEARnet will as well. Must get back to work; we have air raid drills scheduled for 2pm this afternoon and my anti-aircraft gun needs greasing. I'll radio in future reports so long as my lines to Princeton aren't cut. --Kent England, Boston University
morgan@jessica.stanford.edu (RL "Bob" Morgan) (05/30/90)
> Somehow dialup Internet access and SMTP don't go hand and hand in my > mind, my estimate is that your going to have keep a connection open > for about 3 hours every day to have some probablity of synchronizing > with all the SMTP agents pushing mail out of their queues for the > site. Realistically you'll be running uucp/tcp to a site like UUPSI > who is MX'ing for your domain. Indeed, SMTP's assumption that everybody's connected all the time doesn't work well with occasionally-connected hosts. It would seem that the time is ripe for some sort of extension to SMTP to do receiver-initiated transfers to meet this need. Of course you'll still need the MXing host to hold your mail for you. Presumably getting away from uucp is one of the points of all this. Note that POP doesn't make it for the small-business customer. I want my address to be "morgan@mybusiness.com" not "morgan@barrnet.net". I also want to manage my own accounts on my own multi-user system, not ask my provider every time I need a new one. - RL "Bob" Morgan Networking Systems Stanford
barmar@think.com (Barry Margolin) (05/30/90)
In article <1990May29.191125.9800@portia.Stanford.EDU> morgan@jessica.stanford.edu (RL "Bob" Morgan) writes: >Indeed, SMTP's assumption that everybody's connected all the time >doesn't work well with occasionally-connected hosts. It would seem >that the time is ripe for some sort of extension to SMTP to do >receiver-initiated transfers to meet this need. No extension is needed: see the TURN command. However, most SMTP implementation don't enable this command for security reasons. To make it secure you need a way of authenticating the calling SMTP. One possibility would be to allow it only in cases where the name given in the HELO command corresponds to the address being used by the transport layer; however, not all transport layers make this available (what do you do if SMTP is being done over dialup lines?). >Note that POP doesn't make it for the small-business customer. I want >my address to be "morgan@mybusiness.com" not "morgan@barrnet.net". I >also want to manage my own accounts on my own multi-user system, not >ask my provider every time I need a new one. MX records solve the first problem. However, I'm not sure why you need any of this. What's wrong with the MX'ing host periodically trying to connect to the occasionally-connected hosts? Much of the time it will just time out, but when it succeeds it can forward all the accumulated mail. This is the default behavior, so no extensions are necessary. -- Barry Margolin, Thinking Machines Corp. barmar@think.com {uunet,harvard}!think!barmar
amanda@mermaid.intercon.com (Amanda Walker) (05/30/90)
In article <57875@bu.edu.bu.edu>, kwe@bu-it.bu.edu (Kent England) writes: > I see a definite interest on the part of commercial organizations > wishing to join the research and education (R&E) Internet, but I see much > less interest from them in joining a purely commercial internet service- > they don't seem to understand who they would be able to talk to on a purely > commercial internet. Come to think of it, neither do I. In our case, at least, one of the reasons that a connection to what you call the R&E Internet would/will be so valuable is that a lot of the people that we want to communicate with are already there. I suspect that the companies that are currently most interested in the Internet fall into one of two categories: 1. Companies whose customer base is largely already on the Internet. TCP/IP vendors, supercomputer vendors, and so on. We fall into this category, for example. 2. Companies that want to reap the benefits of wide-area networking without having to develop their own infrastructure. Some bicoastal computer corporations fall into this category, for example. As the number of companies in category 2 increase, the perceived value of a purely commercial internet will also increase. At some point, also, commercial service providers may increase its value by allowing it to mediate services that the R&E Internet cannot or by policy will not support. -- Amanda Walker -- This posting is cursed. As you read it you will be confuset by ther printeb wertz. Yer intelijen will vabni ..... XRT! XRT!
craig@bbn.com (Craig Partridge) (05/30/90)
In article <1990May29.191125.9800@portia.Stanford.EDU> morgan@jessica.stanford.edu (RL "Bob" Morgan) writes: > >Indeed, SMTP's assumption that everybody's connected all the time >doesn't work well with occasionally-connected hosts. It would seem >that the time is ripe for some sort of extension to SMTP to do >receiver-initiated transfers to meet this need. Of course you'll >still need the MXing host to hold your mail for you. Presumably >getting away from uucp is one of the points of all this. CSNET did this some time ago with MMDF2b. Some of the dial-up sites run a script every night which brings up the dial-up line, and then opens a connection to a port on relay.cs.net and tells it to start delivering mail to the site. The application at that connection starts up the appropriate MMDF channel (mmdf can have multiple SMTP delivery channels, where a channel typically has messages destined for a particular site), which delivers the mail to the site. [Note there's no security problem here -- anyone can start up the channel, but the channel will only deliver to the proper remote system(s)] Craig
kwe@bu-it.bu.edu (Kent England) (05/31/90)
In article <118@ultrix.uhasun.hartford.edu>, jbloom@uhasun.hartford.edu (Jon Bloom) writes: > In article <57875@bu.edu.bu.edu>, kwe@bu-it.bu.edu (Kent England) writes: > > We also have strict standards on quality of service, and we > > do not wish to compromise these standards in offering less costly access, ... > > > Speaking as one who is trying to figure out how to convince management that > $10k/yr would be well spent, I would be willing to accept service limitations > for a lower-cost net access. If the service truly is as useful to the > organization as I believe it would be, the demonstration of that usefulness > might just break loose the dollars for a higher quality ($10K) connection. So > providing low-cost, restricted service connections may well have the effect > of enhancing the number of sites getting full-service connections eventually. > We hear you. Let me elaborate on some of the costs hidden in a NEARnet service fee. Consider these representative of a mid-level network service provider. I certainly don't mean this as an advertisement, but as explanation of the costs of internetworking. It might be helpful for someone starting from a UUCP or BITnet point of view. Cost of the router on your end. NEARnet requires that the demarcation point be on the local LAN and not somewhere on the leased line. In other words, NEARnet has to own and control the router at your site. You have to pay for that. But this has proved critical to achieving service objectives. Cost of the router on our end. It takes a lot of hardware to make the NEARnet core work and provide access to NSFnet and other services. Leased line charges. Leased lines are expensive and difficult to move and change. Fractional T1 is helping that a little, but still, we are in large part unable to provide distance-insensitive fees, and some of our clients pay large fees because of geography. Depreciation. What does your service provider do for you when your router is obsolete and can no longer function in an Internet? Do they make you pay for a new one? NEARnet builds depreciation into its fee schedules, so you pay for the cost of hardware and software continually, and not in a lump sum, or worse, never. Depreciation is critically important to a network service provider being able to keep up with the technology. People costs. It is very expensive in man-hours to run a mid-level network, but we have found that it does not pay to scrimp on people. I think new network management tools will help, particularly as we gain control of the wide area network, but it is still an expense that we feel is justified to provide a sufficiently high level of service. NOCs should be 24 hours a day, seven days a week, and that will never be cheap. User services. We have found it absolutely critical to be able to offer some minimal user services to our members. This is extremely important to some of our clients. Some of these services are things like name servers for their domains and help with registration and connected-status. But it also includes a committment from NEARnet to take responsibility for tracking down, identifying, and, to the extent possible, fixing users' connectivity problems, no matter where in the Internet the problem is lurking. This is a service available to anyone, anywhere. Finger @nic.near.net for details. It also includes Dan Long's time and effort in the IETF in the User Connectivity Problems Working Group to try to spread this capability and commitment throughout the Internet. All this adds up to quite a chunk of change. Makes UUCP look pretty good, eh? :-) I understand, though, that money is money and it still costs too much. I believe the NSF is sympathetic, as they are still considering proposals for grants to connect qualified sites to the Internet. You may be able to turn to the NSF for grant aid. It's all part of the Internet growing up, running like a business, going private, etc. And NEARnet is not the only mid-level network service provider working these issues. More and more operational people are getting into the IETF and there is the Federation of American Research Networks which is populated by the mid-level networks, among others, and is working on these sorts of issues. --Kent England
schoff@PSI.COM ("Martin Lee Schoffstall") (05/31/90)
Jon, I think the counter argument is that if it doesn't work reliably/well especially in the beginning then people won't rely on it to provide communications for their everyday needs. If they don't use it everyday then it is luxury suitable for axing at the appropriate belt-tightening time. Marty --------------- > Speaking as one who is trying to figure out how to convince management that > $10k/yr would be well spent, I would be willing to accept service limitations > for a lower-cost net access. If the service truly is as useful to the > organization as I believe it would be, the demonstration of that usefulness > might just break loose the dollars for a higher quality ($10K) connection. So > providing low-cost, restricted service connections may well have the effect > of enhancing the number of sites getting full-service connections eventually.
jb.loom@uhasun.UUCP (05/31/90)
In article <57875@bu.edu.bu.edu>, kwe@bu-it.bu.edu (Kent England) writes: > > NEARnet is committed to broadening its base of support to include > smaller sites at lower cost. We are chafing against technical issues in > being able to offer high quality service at lower costs, since we balance > hardware and people costs, and people costs are not really that bandwidth > sensitive. We also have strict standards on quality of service, and we > do not wish to compromise these standards in offering less costly access, > since we know that later on our customers would regret the compromise as > much as we. Then there is the issue of just what is part-time access; > is it terminal dial-ups, SLIP/PPP (host or router?), uucp, what? We have > to keep in mind what services we provide and we have to make sure that > our clients understand what they can and can't do with new service offerings. > Speaking as one who is trying to figure out how to convince management that $10k/yr would be well spent, I would be willing to accept service limitations for a lower-cost net access. If the service truly is as useful to the organization as I believe it would be, the demonstration of that usefulness might just break loose the dollars for a higher quality ($10K) connection. So providing low-cost, restricted service connections may well have the effect of enhancing the number of sites getting full-service connections eventually. Jon -- Jon Bloom, KE3Z -- jbloom@uhasun.hartford.edu
smb@ulysses.att.com (Steven Bellovin) (05/31/90)
There's no reason, of course, why the dial-up hub can't call out. As long as the time to complete the call (including login) is less than 30 seconds or so, most TCPs won't care (too much).
henry@utzoo.uucp (Henry Spencer) (06/01/90)
In article <9005302148.AA02245@psi.com> schoff@PSI.COM ("Martin Lee Schoffstall") writes: >I think the counter argument is that if it doesn't work reliably/well >especially in the beginning then people won't rely on it to provide >communications for their everyday needs... True in general, but one must beware of thinking of "reliably" and "well" as Booleans. Uucp mail is unreliable and ungodly slow by Internet standards, but vast numbers of sites found it reliable *enough* and workable *enough* to come to rely on it very extensively. The reason why a zoology department (!) was a major networking hub at U of T for some years was that everybody else was so obsessed with Internet (or better) performance that they wouldn't even look at silly ideas like networking via low-speed modems. We set up our phone links and started shipping mail and news, and pretty soon half the campus had uucp links to us. Half a loaf is much more nutritious than none. -- As a user I'll take speed over| Henry Spencer at U of Toronto Zoology features any day. -A.Tanenbaum| uunet!attcan!utzoo!henry henry@zoo.toronto.edu
meissner@osf.org (Michael Meissner) (06/01/90)
In article <9005301640.6.142@cup.portal.com> jb.loom@uhasun.UUCP writes: | In article <57875@bu.edu.bu.edu>, kwe@bu-it.bu.edu (Kent England) writes: | > | > NEARnet is committed to broadening its base of support to include | > smaller sites at lower cost. We are chafing against technical issues in | > being able to offer high quality service at lower costs, since we balance | > hardware and people costs, and people costs are not really that bandwidth | > sensitive. We also have strict standards on quality of service, and we | > do not wish to compromise these standards in offering less costly access, | > since we know that later on our customers would regret the compromise as | > much as we. Then there is the issue of just what is part-time access; | > is it terminal dial-ups, SLIP/PPP (host or router?), uucp, what? We have | > to keep in mind what services we provide and we have to make sure that | > our clients understand what they can and can't do with new service offerings. | > | Speaking as one who is trying to figure out how to convince management that | $10k/yr would be well spent, I would be willing to accept service limitations | for a lower-cost net access. If the service truly is as useful to the | organization as I believe it would be, the demonstration of that usefulness | might just break loose the dollars for a higher quality ($10K) connection. So | providing low-cost, restricted service connections may well have the effect | of enhancing the number of sites getting full-service connections eventually. If all you currently want is SMTP and FTP initially, then uunet may be a way to start. They obviously will do mail, but another service they offer to paying customers is to do anonymous FTP's on request, and UUCP it to your machine. When I was at Data General, and the internet connection was extremely flakey (it did get better before I left), I begged, pleaded, and such to management to establish such a UUNET connection and buy the Telebit Trailblazer modem. I was able to get GCC releases and prereleases in a timely manner (though there was the time when I couldn't keep an internet connection or modem connection open long enough, thanks to the lousy GTE phone service in the RTP area of North Carolina or the lousy internal DG phone service). The last time I looked, the uunet charges where $35/month, $2/hour if you called their local number (using company WATTS lines if you have them), or $16/hour if you used their 800 number. A Telebit T2500 modem sells in the range of $1000. Obviously this is subject to change, as is their willingness to do anonymous FTP's for you (though they have most of the stuff people want directly on uunet). -- Michael Meissner email: meissner@osf.org phone: 617-621-8861 Open Software Foundation, 11 Cambridge Center, Cambridge, MA Catproof is an oxymoron, Childproof is nearly so
ddl@husc6.harvard.edu (Dan Lanciani) (06/01/90)
In article <57952@bu.edu.bu.edu>, kwe@bu-it.bu.edu (Kent England) writes: > In article <118@ultrix.uhasun.hartford.edu>, jbloom@uhasun.hartford.edu > (Jon Bloom) writes: > > In article <57875@bu.edu.bu.edu>, kwe@bu-it.bu.edu (Kent England) writes: > > > We also have strict standards on quality of service, and we > > > do not wish to compromise these standards in offering less costly access, > ... > All this adds up to quite a chunk of change. Makes UUCP look pretty > good, eh? :-) One of the reasons that UUCP does indeed look good in this comparison is that the current structure of UUCP networks (last I looked!) still allows a site to join the network at as low a cost and with as poor service as it can tolerate. In many cases, e.g., a local phone call, that service can be pretty good and pretty cheap. Many regional internet service providers forbid "third party" connections and nets-behind-nets because they see such access as depriving them of the revenue they would obtain from a directly-connected customer. While this may be a valid business concern, and while it has the side effect of allowing the regional to enforce a certain quality of service, it can preclude some interesting and potentially cost-effective (at least for the customers) structures. A more standard and quantitative method of charging for internet service, e.g., per-packet accounting (sigh), might allow the existence of low-end sub-service providers without threatening the income of the regionals. That is, a regional need not worry whether two customers pay for 100 packets each or one subcontractor pays for 200. Competition could then drive the price of service down to a point where even an individule can afford a connection... Dan Lanciani ddl@harvard.*
sl@van-bc.UUCP (Stuart Lynne) (06/01/90)
In article <3103@husc6.harvard.edu> ddl@husc6.harvard.edu (Dan Lanciani) writes: >While this may be a valid business concern, and while it has the side effect >of allowing the regional to enforce a certain quality of service, it can >preclude some interesting and potentially cost-effective (at least for the >customers) structures. > A more standard and quantitative method of charging for internet >service, e.g., per-packet accounting (sigh), might allow the existence of >low-end sub-service providers without threatening the income of the regionals. It also provides a way for people to try things out. A sys admin can hook into van-bc for example simply by agreeing to attempt to make some donations to help us run things. So he can hook up, usually with existing resources and then after a few months demonstrate to management that it's a good thing. And then get authorization to do it and make some donations. We also see people using the uucp service for a while and then deciding to move up into SLIP service directly into the regional network. Again if they hadn't had the opportunity to demonstrate the usefullness of the endeavour they might not have been able to get management approval for the more expensive regional network connection. -- Stuart.Lynne@wimsey.bc.ca ubc-cs!van-bc!sl 604-937-7532(voice) 604-939-4768(fax)
medin@NSIPO.NASA.GOV ("Milo S. Medin", NASA ARC NSI Project Office) (06/01/90)
. . . I think the counter argument is that if it doesn't work reliably/well especially in the beginning then people won't rely on it to provide communications for their everyday needs. If they don't use it everyday then it is luxury suitable for axing at the appropriate belt-tightening time. Marty Marty, I think John was trying to espouse the "crack dealer's" approach to Internetting. First, you give them a little something either free or very low in cost. Then, once they are hooked, you stick it to them with high cost, high performance solutions, and by this time, they are screaming for more and can't do without. They will pay and and pay at this point. Those of us who work in the government side of networking don't use this exact tactic, but I know of many large organizations who were "hooked" this way... Thanks, Milo
schoff@PSI.COM ("Martin Lee Schoffstall") (06/01/90)
Many regional internet service providers forbid "third party" connections and nets-behind-nets because they see such access as depriving them of the revenue they would obtain from a directly-connected customer. While this may be a valid business concern, and while it has the side effect of allowing the regional to enforce a certain quality of service, it can preclude some interesting and potentially cost-effective (at least for the customers) structures. Revenue is only one issue, there are considerations of performance, debugging, responsability and market perception: PERFORMANCE: There are cutomers who opt for 9.6-19.2 Internet connectivity, who don't understand why their link is unusable after they feed a couple fortune 500 companies with UUCP behind it, then potentially they want to offload their DNS to their service provider, which most regionals do, but why do we do it, for them or for the companies behind them which we have no relationship with. DEBUGGING: So the MX record doesn't work, so their sendmail.cf is misconfigured in doing this, so the side behind them calls the service provider to ask questions, get help, whatever. For the regionals working with each other all the time, the RIGHT assumption is that everyone is carrying each other at some level, and that the people at the ends are paying for the service. A terminus uucp connection (which is the norm for the problem sites), pays nothing. RESPONSABILITY: Who is responsible for that uucp connection, especially when friend system administrator A and University B moves onto make some big money at Commercial company C. MARKET PERCEPTION: Maybe this is the largest issue of all, without even trying the service quality perception of network Z can be drawn down by a dozen bad uucp connections who's users think they are on ZNet and they tell their friends about how it isn't reliable, doesn't work, poor performance etc... Some organizations including PSI/PSINet throughout the US and CERFNet throughout California have simply decided that the real solution is to establish local dialups everywhere they are, and provide CHEAP UUCP connections and deal with this directly. Marty
karn@ka9q.bellcore.com (Phil Karn) (06/03/90)
In article <1990May29.191125.9800@portia.Stanford.EDU> morgan@jessica.stanford.edu (RL "Bob" Morgan) writes: > >Indeed, SMTP's assumption that everybody's connected all the time >>doesn't work well with occasionally-connected hosts. [...] I'm also very interested in making the Internet protocols work well over non-full-time paths. Here's an idea that I put into my KA9Q TCP/IP package a while ago. It has a UDP-based server that handles "kick" requests. These force retransmissions on any TCP connections to the specified IP address (which defaults to the sender of the kick message) and they also cause the remote SMTP daemon to start sending any pending traffic to the destination. This has turned out to be a useful feature when dealing with dialup links or the long network outages that are characteristic of some experimental amateur packet radio links. When the path is up, the person expecting mail sends a kick request to the site(s) from which he expects his traffic, e.g., a mail exchanger. If there's any traffic waiting, it immediately starts to flow. Also, because my TCP connections never time out (they just back way off) the kick command lets you pick up a partial transfer right where you left off when the path broke without having to restart it from the beginning. Perhaps it's time for an "intermittent connectivity" working group in the IETF? Phil
henry@utzoo.uucp (Henry Spencer) (06/03/90)
In article <36901@think.Think.COM> barmar@nugodot.think.com (Barry Margolin) writes: >>Indeed, SMTP's assumption that everybody's connected all the time >>doesn't work well with occasionally-connected hosts. It would seem >>that the time is ripe for some sort of extension to SMTP... > >No extension is needed: see the TURN command. However, most SMTP >implementation don't enable this command for security reasons... TURN is the wrong tool for the job. What is needed is the equivalent of UUCP's callback facility: A connects to B, says "call me", and hangs up. B then makes the call -- which eliminates the authentication issue -- and transfers the traffic. People interested in infrequently-connected hosts would do well to study how UUCP software and administrators deal with the problems. There is a large body of relevant experience there. >However, I'm not sure why you need any of this. What's wrong with the >MX'ing host periodically trying to connect to the occasionally-connected >hosts? [which is the way it already works] ... It has the standard problem of any polling scheme: frequent attempts eat resources, infrequent attempts can miss "hitting the window" for a site which is only connected occasionally. As with most polling, the real answer is to have the other end *tell* you "now would be a good time". SMTP's approach assumes hosts that are on the network more or less continuously, with partitioning and/or downtime a relatively rare and transient phenomenon that can be dealt with in simplistic ways. When the connection, as opposed to the absence thereof, is rare and transient, more sophisticated tactics are needed. -- As a user I'll take speed over| Henry Spencer at U of Toronto Zoology features any day. -A.Tanenbaum| uunet!attcan!utzoo!henry henry@zoo.toronto.edu
jc@minya.UUCP (John Chambers) (06/04/90)
> CSNET did this some time ago with MMDF2b. Some of the dial-up sites run > a script every night which brings up the dial-up line, and then opens > a connection to a port on relay.cs.net and tells it to start delivering > mail to the site. The application at that connection starts up the > appropriate MMDF channel (mmdf can have multiple SMTP delivery channels, > where a channel typically has messages destined for a particular site), > which delivers the mail to the site. [Note there's no security problem > here -- anyone can start up the channel, but the channel will only deliver > to the proper remote system(s)] How so? What's to prevent me from running a script of the form: for h in foo bar bax glorph `hostname` do hostname $h <<Connect to your machine and exchange mail>> done Is there some mechanism for detecting such spoofing? The only way I can think of is by noting info at the link level (Ethernet address, phone number, or some such) and comparing, but on most systems, this info isn't available to application-level processes. For instance, how does a process started from a modem login discover the caller's phone number? UUCP uses callback (if you can figure out how to make it work ;-), but I didn't see any mention of that. What am I missing? (Yeah, I know I'll also have to run ifconfig for this to work across the internet. That's part of the <<code>; give me credit for some smarts! :-) -- John Chambers ...!{harvard,ima,mit-eddie}!minya!jc -- If you're not part of the solution, you're part of the precipitate. [Kiel oni ^ci tiun diras esperante?]
wcs) (06/06/90)
In article <1990May31.211414.8140@utzoo.uucp> henry@utzoo.uucp (Henry Spencer) writes:
]The reason why a zoology department (!) was a major networking hub
]at U of T for some years was that everybody else was so obsessed with
]Internet (or better) performance that they wouldn't even look at silly
]ideas like networking via low-speed modems.
]--
]As a user I'll take speed over| Henry Spencer at U of Toronto Zoology
]features any day. -A.Tanenbaum| uunet!attcan!utzoo!henry henry@zoo.toronto.edu
It's nice to see that Henry does have a balanced view of things,
regardless of what his .signature may imply :-) (Tanenbaum, aside
from any OSI politics, is the author of Minix, which was designed to
provide features at the cost of some speed on a machine where speed
was in short supply already - a successful and nice project.)
I've always been surprised at how much you could accomplish with UUCP;
my main disagreement with the SMTP approach, other than the
appalling overcomplexity of sendmail, is that it wants to make a
connection NOW, and send traffic the whole way, rather than settling
for store-and-foreward on a message basis. While uucp was always a
pain to maintain, the major cause of problems seemed to be
inadequate numbers of modem ports.
--
Thanks; Bill
# Bill Stewart AT&T Bell Labs 4M312 Holmdel NJ 201-949-0705 erebus.att.com!wcs
# Actually, it's *two* drummers, and we're not marching, we're *dancing*.
# But that's the general idea.
david@twg.com (David S. Herron) (06/06/90)
In article <9005270423.AA19852@psi.com> schoff@PSI.COM ("Martin Lee Schoffstall") writes: > The plight of many small technical businesses is that we just cannot > justify spending $30K+... > Were access fees brought inline with the level of > service offered >Somehow dialup Internet access and SMTP don't go hand and hand in my mind, >my estimate is that your going to have keep a connection open for about 3 hours >every day to have some probablity of synchronizing with all the SMTP >agents pushing mail out of their queues for the site. Realistically you'll >be running uucp/tcp to a site like UUPSI who is MX'ing for your domain. Hmm.. interesting discussion. uucp/tcp isn't a good choice because the transition through rmail loses information and bursts messages going to multiple recipients. There's a bunch of CSNET like tricks that can be played to make this work out better. For instance, a "relay" machine could be advertised as a fairly low cost MX alternative for the true site causing mail to be dropped at the relay. Oh, but does sendmail choke a bit when it has lots of undeliverable mail in its queue? Sigh, such is life... :-) Count me in as a potential customer of low cost IP service. It'd be real nice, when travelling, to be able to reach my home site w/o problem. For day-day work-at-home stuff I think I can manage to arrange something local without too much expense, especially since my home line isn't charged as a business line. "On the road" is another matter entirely.. -- <- David Herron, an MMDF weenie, <david@twg.com> <- Formerly: David Herron -- NonResident E-Mail Hack <david@ms.uky.edu> <- <- Sign me up for one "I survived Jaka's Story" T-shirt!
wuu@DUMBO.JVNC.NET (Sze-ying Wuu) (06/12/90)
> CSNET did this some time ago with MMDF2b. Some of the dial-up sites run > a script every night which brings up the dial-up line, and then opens > a connection to a port on relay.cs.net and tells it to start delivering > mail to the site. The application at that connection starts up the > appropriate MMDF channel (mmdf can have multiple SMTP delivery channels, > where a channel typically has messages destined for a particular site), > which delivers the mail to the site. [Note there's no security problem > here -- anyone can start up the channel, but the channel will only deliver > to the proper remote system(s)] > > Craig > Does CSNET uses Phonenet protocol for dial-up sites? We have 1986 version of MMDF, is there a newer version available from pulic domain? I assume with MMDF, we still need MX RR for a host to accumulate mail for all dial-up sites. Is there a limitation on the number of channnels can be fired up? Sze-Ying Wuu JvNCnet wuu@nisc.jvnc.net