Carl Malamud: Internet Talk Radio, flame of the Internet. 

This is Geek of the Week and we’re talk­ing with Peter Deutsch, who is pres­i­dent of Bunyip Information Systems. Peter, wel­come to Geek of the Week.

Peter Deutsch: Glad to be here.

Malamud: You’re best known as the orig­i­na­tor of Archie, or one of the orig­i­na­tors of Archie. Can you tell us what Archie is?

Deutsch: Okay, for those who have been on the net all that long or hav­ing a chance to find Archie, what we’ve basi­cal­ly got is a tool that goes out across the Internet, col­lects infor­ma­tion, and puts it all in one place so that you don’t have to know where every­thing is, you just have to know where we are and we’ll help you find it.

Malamud: So you go out and do direc­to­ry list­ings of anony­mous FTP sources.

Deutsch: Yeah, that’s our cur­rent famous data­base. We cur­rent­ly offer two, but the most famous one is a list of basi­cal­ly all the anony­mous FTP archive sites on the net that will allow us to track them. It’s a vol­un­teer thing. If you don’t want us, you tell us, we won’t track you.

Malamud: Now how did you end up start­ing Archie? Why’d you do some­thing like this?

Deutsch: Okay. Well, I was a grad­u­ate stu­dent work­ing for the School of Computer Science at McGill. I had a cou­ple of peo­ple work­ing for me. The way it usu­al­ly works if you work for a uni­ver­si­ty is you work on the ten-year plan. They allow you to work, they swear it does­n’t inter­fere with your stud­ies. And…it does. 

So what we had there was a fair­ly large num­ber of machines. We had an Internet link and we have to keep it alive with no mon­ey for soft­ware or any­thing like that. So my part­ner Alan Emtage, who worked for me at the time, was giv­en the job of find­ing the miss­ing things. We know there’s some­thing called gnu­tar, where is it? Alan’s lazy. He likes to think of him­self as a true com­put­er sci­en­tist, that is to say he min­i­mizes key­strokes. So he wrote this lit­tle tool. And unfor­tu­nate­ly we men­tioned it in a Usenet post­ing once, got flood­ed with post­ings with peo­ple ask­ing us to do search­es. So we stuck fron­tend onto it, and that was sort of it. It was off and run­ning at that point.

Malamud: Well you start­ed it off by run­ning one Archie serv­er at McGill. And then all of a sud­den there began to be lots of Archie servers. There’s one in Australia, there’s one in Finland. That means every one of these Archie servers goes out to these anony­mous FTP sites and does a direc­to­ry list­ing. And peo­ple look at Archie and say well gee, one Archie is fine. But what hap­pens when a mil­lion Archies come in every night and do my direc­to­ry list?

Deutsch: Okay, we were actu­al­ly very aware of this from the begin­ning. We were very aware of the load that Archie start­ed to gen­er­ate at McGill, because it is a resource-intensive ser­vice. We had a lot of peo­ple com­ing to use it. So we looked to see about spread­ing it around. Now, we could’ve just stuck the code in archive, but we said there’s a prob­lem with that. And the prob­lem is ten mil­lion Archies hit­ting all the archives would be a problem. 

So what we decid­ed to do is we iden­ti­fied some peo­ple who would help us sort of coop­er­ate. We worked with FUNET and AARNET in Australia, and ANS and SURAnet and so on. And we’ve set it up so that the var­i­ous Archies actu­al­ly talk to each oth­er. So one Archie will gath­er all the files in Sweden, for exam­ple. And then rather than all the oth­er Archies com­ing to vis­it Sweden, they will just con­tact this one, he will send a copy, who will send a copy, and so on. There’s a fair amount of hand­shak­ing going on behind the scenes that the user does­n’t see. And it helps to min­i­mize the loads.

Malamud: This kind of sounds like a glob­al direc­to­ry, does­n’t it?

Deutsch: Uh…under a dif­fer­ent name? Sure. It hap­pens to track only anony­mous FTP files, at this moment, although we do hope in the rea­son­ably short term that there’ll be more col­lec­tions served by this. But, yeah it’s a direc­to­ry ser­vice for anony­mous FTP. We like to think the tech­nol­o­gy is exten­si­ble and can be used for oth­er col­lec­tions. And as soon as we can get a lit­tle bit more time to stop answer­ing mail and start build­ing, we’ll have a few more data­bas­es out there.

Malamud: So is Archie X.500-compliant?

Deutsch: [laughs] Okay you found my hot but­ton. You’re look­ing for a flame and you’re gonna have one. What we did is… We need­ed some direc­to­ry ser­vices. And we looked around and said okay, let’s look at this X.500 stuff. And we brought it in and we played with it for a cou­ple of days. And like many peo­ple, unfor­tu­nate­ly, we threw it away. It’s too hard to bring up. I think they’ve con­fused imple­men­ta­tion with archi­tec­ture a few times, and it basi­cal­ly did­n’t do what I want­ed it to do. 

So, Archie is an inter­est­ing col­lec­tion of infor­ma­tion. It does­n’t every­thing stan­dard­ized in terms of how you can address queries to it. So there is anoth­er effort we have going on at the IETF to build a sort of very light­weight, sim­ple pro­to­col for doing direc­to­ry queries. What we would like to see is a com­bi­na­tion of Archie, which gath­ers infor­ma­tion into one place, and this new pro­to­col we’re work­ing on, an enhance­ment WHOIS we called WHOIS++. And I think that’s going to be a fair­ly inter­est­ing com­bi­na­tion when we final­ly fin­ish it off.

Malamud: Well, what’s wrong with the X.500? There’s been a lot of work that’s gone into it. It’s gone through a cou­ple iter­a­tions in the inter­na­tion­al stan­dards effort. It’s tak­en in input from all par­ties. Isn’t that the way we ought to be doing directories?

Deutsch: I’ve spo­ken with Steve Hardcastle-Kille, one of the peo­ple behind it, and he told me that we should use X.500 because it’s the best polit­i­cal solu­tion. Now, the prob­lem I have with that is I don’t think the Internet is built on the best polit­i­cal solu­tions, I want things built on the best tech­ni­cal solu­tions. It is true that var­i­ous polit­i­cal bod­ies have had input into this. I don’t see that a lot of users have had a lot of input into this. And I don’t see that there’s a lot of use­ful infor­ma­tion in there. I would like some­thing that’s dri­ven by demand, and…some— My own par­tic­u­lar view on things—maybe I’m a fair­ly sim­ple per­son, but I like to build fair­ly sim­ple things and make it incre­men­tal­ly more com­plex until it does just enough, and no more. And I’m afraid that they’ve sort of thr…they’ve thrown an awful lot at this problem.

Malamud: It’s feature-rich.

Deutsch: It’s feature-rich, boy is it. And I would rather take a step— There were some very inter­est­ing things. I don’t want to crit­i­cize the peo­ple that’ve done this work, because some very inter­est­ing ideas have come out of it. But there’s also some things we’ve learned, that’ve been done wrong, and I think we can go back and take anoth­er slice at it. And I’m not say­ing we have to throw X.500 away. There’s been a fair amount of effort put into it, we can cer­tain­ly improve it, and hope­ful­ly the work we do with WHOIS…if noth­ing else it encour­ages X.500 to fix some things it’s got wrong with it [all would have don a use?]. I have no pro­pri­etary inter­est in you know, get­ting WHOIS out there, I just want a work­ing direc­to­ry ser­vice, and if no one else will give me one that works that I can just turn on out of the box, I’m gonna help build one myself.

Malamud: Now, what is WHOIS++? How is this dif­fer­ent than WHOIS?

Deutsch: Okay, what we’ve basi­cal­ly done here is… Our man­date, which we set our­selves a cou­ple of months ago, or a few months ago, was we want to be able to answer quite sim­ple ques­tions. What is your email address? What is your name? Where is the near­est gopher serv­er and so on. So, what we did is we looked at WHOIS as a mod­el. There was a BoF held late in 1992 at an IETF, and we kicked some ideas around. And what came out of it was well no, we should­n’t try to do X.500 all over again. That’d def­i­nite­ly be inap­pro­pri­ate. But would it be appro­pri­ate to look at enhanc­ing the sim­ple WHOIS mod­el. WHOIS in effect has a pro­to­col, if you will: send me some­thing, I’ll send you some­thing back. And what we’re doing is we’re jazz­ing that up a lit­tle bit: send me some­thing a lit­tle struc­tured, I’ll send you some­thing back a lit­tle structured. 

So, it’s an enhance­ment to WHOIS where we basi­cal­ly have tak­en some work that’s been done at the IETF on defin­ing stan­dard­ized tem­plates for encod­ing infor­ma­tion, and we’re say­ing okay, now let’s have a mod­el that says ​“I have a col­lec­tion of these tem­plates and I can search on attrib­ut­es, and I can search on val­ues, and I can give you back the results. And it’s—we’re try­ing to keep it as sim­ple as pos­si­ble. We don’t want to do every­thing on the first pass. What we want to do is be able to answer some sim­ple ques­tions like ​“What is Carl Malamud’s email address?” you know, that should not be brain surgery, and right now it’s one of the tougher prob­lems on the net. 

Malamud: So what’s wrong with SQL and these oth­er lan­guages that define query mech­a­nisms and response mechanisms?

Deutsch: Okay, well I out­lined what we were doing with WHOIS++ to some­body and he said, ​“Well you know, I have a pro­to­col that would be per­fect for this appli­ca­tion. You don’t have to use your own pro­to­col, which is real­ly quite sim­ple. You can map every­thing you have into one of my fea­tures in my pro­to­col. And of course my polit­i­cal does a gazil­lion oth­er things as well.” 

And imme­di­ate­ly a lit­tle alarm bell goes off. And the prob­lem is, we real­ly want to mod­el our­selves after the suc­cess of Gopher. The Gopher peo­ple have a very sim­ple query trans­ac­tion pro­to­col: send some­thing that looks like this, we’ll send you back some­thing that looks like this. And because it was so sim­ple, a lot of peo­ple can exper­i­ment with it. They could build clients, they could build servers, they could bring in inter­est­ing vari­a­tions to it, because build­ing the pro­to­col stack is no big deal. You can spend your time on the enhance­ments and the fea­tures, because that’s where we want to see the exper­i­men­ta­tion. I don’t think it should be that hard to ask sim­ple questions.

Now, if you need com­pli­cat­ed ques­tions, fine. There are more com­pli­cat­ed solu­tions. But the way I describe it to peo­ple is, if you think direc­to­ry ser­vices is a hard prob­lem you’ll have a hard solu­tion. Let’s go for some sim­ple solu­tions, let’s ask some sim­ple ques­tions, and see what comes out of it. So that’s basi­cal­ly were takin’ it right now.

Malamud: You work in a gen­er­al area known as resource dis­cov­ery, and there’s a vari­ety of efforts out there in the resource dis­cov­ery realm. Are all these gonna come togeth­er and become one sin­gle ser­vice, or are they gonna work together…what’s gonna hap­pen? Is Archie gonna merge with WAIS?

Deutsch: I— This is inter­est­ing. There’s a cou­ple of ways you can take this, or you can try to take this. And one attempt could be made to find the pro­to­col to end all pro­to­cols, the ser­vice to end all ser­vices, and tie it all togeth­er. I think that’s inap­pro­pri­ate for a num­ber of rea­sons. Amongst oth­er things if you look at the way inter­est­ing and use­ful or pro­duc­tive things have been built on the net—the whole net has been built—it’s been a slice at a time. We don’t eat our sala­mi by tak­ing the whole thing and try­ing to chew it. We take a slice of the prob­lem at a time. 

And you don’t want to dis­cour­age that. If you try to go for the one glob­al solu­tion I think you’re gonna have…you know, right back where we were talk­ing a few min­utes ago, the prob­lem that tried to do too much. So I think what you’re gonna find is, there will be an inte­gra­tion effort, cer­tain­ly we’ve seen it already. I am quite hap­py to say that Archie is not real­ly an end user client ser­vice. It real­ly should be hid­den behind some­thing like Gopher or World Wide Web. These peo­ple have spent time think­ing about the inter­ac­tion prob­lem, and they’ve got a good cut, pre­sent­ing infor­ma­tion to users in a way that’s fair­ly easy to do. 

Now, I want Archie and I want WAIS, and these oth­er index­ing ser­vices hid­den behind a nice front end. I want the peo­ple who do fron­tends well to con­cen­trate on them. There will be an inte­gra­tion in that sense. But what I don’t think you’re gonna see is Archie won’t become Gopher or vice ver­sa. We will have inter­est­ing col­lec­tions, and if you want the inter­est­ing col­lec­tion you’ll learn to speak the access method you need for that. We’ll still have lots of them but I think the tools will become…we’ve got a multi-protocol net­work, we’ll have a multi-protocol infor­ma­tion ser­vice as well. 

Malamud: You’re lis­ten­ing to Geek of the Week. Support for this pro­gram is pro­vid­ed by O’Reilly & Associates, rec­og­nized world­wide for defin­i­tive books on the Internet, Unix, the X Window System, and oth­er tech­ni­cal top­ics. Additional sup­port for Geek of the Week comes from Sun Microsystems. Sun, the Network is the Computer.

Malamud: Now, you men­tioned Gopher and World Wide Web. And in a way those are two extremes of how to do an imple­men­ta­tion. Gopher is very sim­ple, almost spare. The pro­to­col’s sim­ple, the user inter­face is fair­ly sim­ple. World Wide Web makes exten­sive use of SGML, of hyper­text links, and might even be hard to put infor­ma­tion into. Certainly once you put the infor­ma­tion in it yields a lot of results. Are both those approach­es gonna be on the net­work or do you think one of those is going to win out?

Deutsch: No— Once again, I keep talk­ing about Darwinian selec­tion of ser­vices, Darwinian selec­tion of pro­to­cols. What you’re going to find is we’re gonna find eco­log­i­cal nich­es where each belongs. You need com­plex solu­tions when you have com­plex prob­lems. What you need is to not try to throw a com­plex solu­tion at a sim­ple prob­lem. Gopher, hope­ful­ly the guys—and from what I’ve seen so far they’re doing an excel­lent job of remem­ber­ing their roots, where they came from, sim­ple is good. They can’t do every­thing but they’re not try­ing to do everything. 

On the oth­er hand, the World Wide Web project…it does have val­ues added. Now, I think what we have to do is iden­ti­fy when is it appro­pri­ate to use some­thing like World Wide Web, which is a hyper­text sys­tem. Very nice­ly dis­trib­uted one, but it’s a hyper­text sys­tem and it has all the asso­ci­at­ed nav­i­ga­tion prob­lems that there is with hypertext. 

So I think putting World Wide Web into a closed domain of infor­ma­tion where you can browse around it is an excel­lent idea. Putting the whole world onto hyper­text I think is a bad idea. So I guess you’ll have it, you’ll have both of them, and you’ll prob­a­bly have some new ones that hope­ful­ly will come out in the next lit­tle while, and I want to see peo­ple exper­i­ment­ing with new ideas. We cer­tain­ly haven’t dis­cov­ered every­thing there is to dis­cov­er about how to do this. But no, no one is going to win out. You gonna have mul­ti­ple pro­to­cols, mul­ti­ple ways of think­ing about infor­ma­tion and there­fore mul­ti­ple ways of surf­ing it.

Malamud: You’ve been look­ing at a vari­ety of ways of iden­ti­fy­ing infor­ma­tion. We have these huge data archives, and you can find a file name but it turns out that the process is more dif­fi­cult because we have more than files we’re try­ing to find and they’re locat­ed all over the place. Can you tell us how we’re gonna iden­ti­fy resources on a network?

Deutsch: Sure. Actually, what we did is my part­ner, Alan Emtage, and I got Archie up and run­ning, and we began to real­ize that you know, Archie can be used for oth­er things, not just the file names. In effect what you’re doing with Archie is you’re search­ing on a par­tic­u­lar attribute, which hap­pens to be the file name. Now, the obvi­ous exten­sion is let’s give it some more attrib­ut­es that we can cat­a­log, and there­fore you can do some more sort of…more gen­er­al searching. 

So, what we did is we cochaired and IETF work­ing group that’s just come to com­ple­tion that’s defined a set of tem­plates that you know, if you want to use one of these Internet— I’ll give you the— The buzz­word here is an IAFA tem­plate. Internet Anonymous FTP Archive was the name of the work­ing group. So we talked about these IAFA tem­plates. There is a user tem­plate that lists var­i­ous attrib­ut­es about a user. His name, his email address­es, his paper mail address­es, his phone num­bers, and so on. Now, this will be an infor­ma­tion­al RFC. It’s just going into draft form right now. And the idea’s we are going to strong­ly encour­age anony­mous FTP archive admin­is­tra­tors to put out col­lec­tions of infor­ma­tion that use these tem­plates, because tools will then be able to know the for­mat of the infor­ma­tion, pick ​’em up, index ​’em, and serve ​’em back out again. We’re not the only peo­ple look­ing to do this, but cer­tain­ly we’re com­mit­ted, as this infor­ma­tion starts to become avail­able, build­ing addi­tion­al data­bas­es. How about one that lists every ser­vice at your site? Well we picked that up from all the sites, we all of a sud­den have solved the yel­low pages prob­lem. We can build services— 

Malamud: So these IAFA tem­plates are cat­a­logs in a card cat­a­log, is that what they are?

Deutsch: In effect what we defined is a stan­dard set of card cat­a­log types that says okay, if you wan­na cat­a­log a per­son, here’s one way to do it. Now, you’re not required to use this way, but if you do and every­one does, then all of a sud­den the tools can be a lit­tle more clever about what they do. So yeah, they’re basi­cal­ly sets— They’re…I can’t remem­ber the exact num­ber but there was about a half dozen types we did. One that spe­cial­izes in users, one that spe­cial­izes in ser­vices, one that describes soft­ware, one that’s an abstract for papers. And you basi­cal­ly cat­a­log what­ev­er you want to be cat­a­logued. Let the peo­ple who’re gonna run these index­ing ser­vices know about it. They’ll start pick­ing them up. If you make a change, the changes will appear auto­mat­i­cal­ly. It’s the Archie idea car­ried to the next step. Other kinds of infor­ma­tion. We all need infor­ma­tion on the net. Nobody wants raw con­nec­tiv­i­ty, they want the infor­ma­tion, and this is our cut at try­ing to get it out there in a way that every­body does a lit­tle slice of it, only the infor­ma­tion of inter­est to you, you keep yours up to date, and then the tools can be used to help every­body find it.

Malamud: Now, I see a big dif­fer­ence between Archie and the IAFA. Archie depends on com­put­ers to get its infor­ma­tion and IAFA depends on peo­ple. What makes you think peo­ple are gonna keep infor­ma­tion up to date?

Deutsch: People often come to me, even when I was still work­ing at McGill University as an admin­is­tra­tor. And the big debate has been in this envi­ron­ment of ​“Oh, we can’t pos­si­bly let the users have access to their own data. We have to con­trol the data.” And I’ve always thought that was real­ly a sil­ly dis­tinc­tion. The peo­ple who run the machines should run the machines, but if I— Let’s face it, if I want my phone num­ber on the net, I want it to be the right phone num­ber. And I’m appro­pri­ate per­son to keep track of that infor­ma­tion and keep it up to date. It’s expen­sive for a com­put­ing cen­ter to pay some­one to keep every­body’s phone num­ber up to date. It’s cheap for each per­son to keep his own num­ber up to date.

Malamud: Well but can peo­ple be expect­ed to— You know, a phone num­ber’s sim­ple. You either know it or you don’t. But when it comes to cat­a­loging the tech­ni­cal papers that’re avail­able and your expect­ed [crosstalk] to come up with codices— 

Deutsch: Oh, I don’t nec­es­sar­i­ly believe that every per­son will be his own librar­i­an and do his own abstracts. But it’s cer­tain­ly appro­pri­ate for exam­ple that your depart­ment has some­one that does this for the whole depart­ment. Now, that’s not out­landish. You could have some­one do that part-time. Trying to do this on the scale of an entire net­work, or an entire uni­ver­si­ty, it gets hard­er. And each time you try to be more ambi­tious, it gets hard­er. And there’s these…reverse economies of scale, if you will, where it’s easy to do a few, it’s hard to do a lot and get oth­er work done. So our own par­tic­u­lar cut on this is that if we can just…you know, the sala­mi mod­el. Take a lit­tle slice. If every­one takes a slice, well very quick­ly the work is done. After all, you have an inter­est that your infor­ma­tion stays up to date. And if you have the tools to keep it up to date…someone com­plains, you’ll fix it. I think it’s a much bet­ter mod­el than try­ing to go for glob­al admin­is­tra­tion. There are times you need it. But…there are times when it’s inap­pro­pri­ate. It’s just not gonna work because it’s too expen­sive on a net­work of over a mil­lion machines.

Malamud: You’re try­ing to define some stan­dards. You’re try­ing to define stan­dards for tem­plate, you’re doing work in areas like uni­ver­sal record iden­ti­fiers. What’s the prop­er place to get those stan­dards done so that they actu­al­ly take effect?

Deutsch: You know, I— This is going to turn into an ad for the IETF. I’m a real enthu­si­ast for the IETF. Every year there’s some ker­fuf­fle that comes up and there’s a lit­tle cri­sis, and I look at this and I think you know, if we have one major fight in the IETF a year, we’re doing pret­ty good, because look how much work is get­ting done. I think it’s an excel­lent mod­el for any oth­er group that wants to set stan­dards. You build your rough con­sen­sus and your work­ing code, and so much work gets done through this avenue. 

So, we’ve con­cen­trat­ed around we’re a fair­ly small com­pa­ny now. We have to be very care­ful about where we you know, spend our ammu­ni­tion. And we’re very com­mit­ted to work­ing through the IETF because I’ve found that it’s a process that pro­duces results quick­ly. There’s a moti­va­tion to get on with it, the only peo­ple there are the ones who actu­al­ly want to work on the prob­lem. It’s democ­ra­cy in action and I think it works pret­ty well. So to me the avenue to try to stan­dard­ize these ser­vices is not to dic­tate a pro­to­col. It’s not to get gov­ern­ments to fund stud­ies. It’s to get vol­un­teers who want the prob­lem solved at the IETF and set up a work­ing group and get some work done.

Malamud: And you think these infor­mal de fac­to stan­dards have a chance of being adopt­ed for these very large data archives out there?

Deutsch: We have an exis­tence proof of one and a half mil­lion machines that talk to each oth­er. The Internet…just works. I mean…it does­n’t always work well, and pieces of it can hurt a lit­tle bit. But I think what’s been accom­plished on the Internet work­ing through this mech­a­nism is pret­ty impres­sive, and I have some faith that it’s got a lit­tle bit life left in it yet.

Malamud: Archie worked, many would say, because it was a free ser­vice. It was avail­able on the net­work, the soft­ware was avail­able, you were run­ning Archie as a non-commercial, any­one could tel­net in and grab their infor­ma­tion. And now you’ve formed Archie Inc, if you will, Bunyip Information Systems. Are you now mov­ing away from that free­ware models?

Deutsch: No. Well, one thing I want to make clear is Archie was nev­er free. It cost, orig­i­nal­ly McGill University a huge amount of resources. We did all of our work, vol­un­teers, as stu­dents. Alan got a mas­ter’s project and I got a the­sis out of it. 

But, to me, we must make users aware that there is a dif­fer­ence between freely avail­able and free. Archie has always been freely avail­able. I hope that that par­tic­u­lar col­lec­tion of infor­ma­tion will always con­tin­ue to be freely avail­able. But when it became clear that we could not work on it at McGill University and stay at McGill University once we fin­ished our stud­ies, then we have to find anoth­er way to pay ourselves. 

So what we cur­rent­ly do is we charge for access to the Archie serv­er code. And that allows us to con­tin­ue to improve it, to bring up new data­bas­es, to go to the IETF. This is fair­ly sim­ple Economics 101 stuff, but…I do have to eat. I’m not inde­pen­dent­ly wealthy. I’m not even depen­dent­ly wealthy. So, I have to have some income. 

Now, there is a lot­ta debate about fund­ing mod­els on the net. I cer­tain­ly am averse to charg­ing some­body ten cents a query. For some very sim­ple rea­sons. Not the least of which is a very effi­cient way of doing it. I know some­body that just got a bill for $2 from AT&T for some phone ser­vice and I thought oh, that’s real­ly sil­ly. Probably cost them four times that to send the bill, if not more. 

So I cer­tain­ly hope that we can find fund­ing mod­els where, for exam­ple, if your net­work is using a ser­vice, your net­work pays the ser­vice provider or the infor­ma­tion provider some flat feet. We are all active­ly look­ing at this. There are a lot of peo­ple that want to put a lot of infor­ma­tion on the net, but we do have to fig­ure out some way to reward this effort. Because you nev­er know what you don’t get. But I can say that, quite frankly if I’d had some sup­port, if I’d had some mon­ey two, three years ago, Archie would be a much more use­ful col­lec­tion of infor­ma­tion than it is now. Now we’ve final­ly got the mon­ey flow­ing, it will get more useful.

So I think that we mi— And there’s con­tro­ver­sy going on over Gopher right now. The Gopher peo­ple have announced that they’re gonna have to start charg­ing license fees as well. And obvi­ous­ly I…you know, I’ve bought into the mod­el where we have to do this, but I can only defend them and say the Gopher peo­ple have shown us some won­der­ful things and they’ve shown what we can do. But they too have pres­sures. Their insti­tu­tion has to be able to see some return on all the efforts they’ve put into it. They’re not out to make a for­tune, we’re not out to make a for­tune. But we do need to get some mon­ey going back to the sources of information—or you just not gonna have ​’em on the net. It’s as sim­ple as that.

Malamud: This is Geek of the Week, fea­tur­ing inter­views with promi­nent mem­bers of the tech­ni­cal com­mu­ni­ty. Geek of the Week is brought to you by O’Reilly & Associates, and by Sun Microsystems.

Malamud: Would Archie had become pop­u­lar if you had not giv­en the way the serv­er code? Could you have sold serv­er code for Archie when you first started?

Deutsch: There’s cer­tain­ly an evo­lu­tion going on on the net. If we’d gone to the peo­ple orig­i­nal­ly and said, ​“We have this real­ly spiffy idea called Archie, and it’s up at McGill and it’s flood­ing our net­work, and I’m gonna sell it to you,” most peo­ple, if not every­body, prob­a­bly would have said, ​“Go away. Don’t both­er me. Don’t be sil­ly. We don’t charge for soft­ware on the net.”

Now we go to them after it’s been up and run­ning for a while. It’s a demon­stra­bly use­ful ser­vice. It’s…you know, not to put too fine a point on it, I don’t want to sort of boast, but it is a use­ful ser­vice that get a lot of use every day. And so the peo­ple that are run­ning it, when we came to them and final­ly said, ​“Look, you’ve now got it, you’ve been run­ning it for a while. If you want us to con­tin­ue to sup­port it you’ll have to help,” almost all of them turned around imme­di­ate­ly and said, ​“Sure, no prob­lem. We understand.”

Now, again we’re not charg­ing thou­sands and thou­sands and thou­sands of dol­lars. However we are cov­er­ing expens­es right now, and— [crosstalk] The answer is—

Malamud: Doesn’t that vio­late the spir­it of the net, though?

Deutsch: Well okay, I guess I did­n’t answer the ques­tion, I gave a lit­tle speech. So, I think what you’re see­ing is the spir­it of the net…has to evolve a lit­tle bit. The spir­it of the net is still there. I still am giv­ing my own back to the net. I go to the IETF and no one pays me to do that. I have work to sup­port, and I go to con­fer­ences and give talks and so on that I don’t charge for. But there is… Let’s face it, what could be done on an Internet of 10,000 machines or a hun­dred thou­sand machines becomes infea­si­ble on an Internet of a mil­lion machines. A good exam­ple of that is Veronica, which is an Archie-like ser­vice that came out for Gopher. Tried to index all of Gopherspace and make that infor­ma­tion avail­able. And I watched Veronica go through in a mat­ter of a few weeks what took us a year to go through. That peri­od of inter­est, more inter­est, explo­sive growth, floods that have to get forked off to oth­er sites, what’re we gonna do about sup­port­ing it, doc­u­men­ta­tion… You know, there’s a whole life­time, and what’s hap­pened now with the size of the net and the pop­u­lar­i­ty of ser­vices, that life­time’s been con­densed. What took us a year to get to the point where it was insup­port­able, took Veronica the space of a cou­ple weeks to become insup­port­able. What used to work on the hun­dred thousand-node net, we’re gonna have trou­ble doing it the same way. 

Now that’s not to say I don’t want to see anoth­er Archie. I most def­i­nite­ly don’t want to dis­cour­age exper­i­men­ta­tion. But one we can do this is take a step back and say okay, how can we cre­ate an envi­ron­ment where we can still allow these pilot ser­vices to be brought up, but without…quite frankly, melt­ing the wire to your uni­ver­si­ty or your research cen­ter or what­ev­er when you first turn it on. I don’t know the answer to that. I only know that it’s a prob­lem we’re going to have to address. And I have every faith that the prob­lems will be addressed because there’s a hack­er feel­ing out there that’s going to con­tin­ue and no one’s gonna kill it. It’s just a ques­tion of okay, the rules have changed a lit­tle bit, a lit­tle bit of envi­ron­men­tal change or Darwinian selec­tion will per­haps find a slight­ly dif­fer­ent way of arriv­ing at the same ends. The hack­ers aren’t gonna stop hack­ing, you can be sure that.

Malamud: You talk about vol­un­teer­ing your time to the IETF. Is that process gonna con­tin­ue or are we gonna have to fund the IETF through some kind of an IETF tax?

Deutsch: Interesting— There’s been a debate going on the last cou­ple years, the IETF’s had the same explo­sive growth that the rest of the Internet has had. My cut on it is that it still works. I think the process is still best served by peo­ple com­ing there as indi­vid­u­als and not as rep­re­sen­ta­tives of spe­cial inter­est groups or par­tic­u­lar slices of the com­mu­ni­ty. My own expec­ta­tion, and my own hope, is that it con­tin­ues the way it is now. Whether that’s true in ten years, I can’t say. I think any­one who tries to pre­dict out more than a year or two in this busi­ness is crazy. But cer­tain­ly I think in the next year two— There are prob­lems to be addressed, no ques­tions about it. But I think that yes, vol­un­teers is still the way to go for at least the next year or so.

Malamud: Well but the fed­er­al gov­ern­ment has been pitch­ing in a good mil­lion dol­lars a year to make the IETF work. And if that mon­ey goes away, how does a process like the IETF hap­pen? Do we raise user fees? Do we raise fees on atten­dees? Do we put a tax on TCP/IP soft­ware? What’s the right way to fund this process of mov­ing tech­nol­o­gy out into the Internet?

Deutsch: Good ques­tion. Alright. Um… There’s a per­ceived val­ue in the IETF. People are very hap­py with what comes out of it. People are very hap­py with the pro— Well. Not every­body’s hap­py with the process, but I think peo­ple are usu­al­ly hap­py with the results. 

I think it would be fair to expect spon­sors of the IETF to sort of pick up the slack if the gov­ern­ment has to cut back. Now, I don’t nec­es­sar­i­ly see the gov­ern­ment has to cut back, giv­en the talk of com­mit­ment to net­work­ing. But let’s assume the gov­ern­ment did cut back. I think it would be appro­pri­ate to accept dona­tions from com­pa­nies that see a ben­e­fit, pro­vid­ed we could make sure that this is strict­ly hands-off dona­tions and you know, there’s a pot you pay into if you think it’s a good ser­vice. Hey, I’ve got no mon­ey, I’m a very small start­up. But I would cer­tain­ly pay larg­er fees if that’s what it took. 

Now, the oth­er thing we might have to look at, the IETF is get­ting very large. There’s been a lot of talk of break­ing it into small­er pieces. If they’re small­er pieces they would­n’t be quite expen­sive on a per-group basis, although you’d lose the…you know, the cheap dough­nuts and the cheap cof­fee because it all has to be done for everybody.

Let’s just take the worst sce­nario. There is a cut­back. I’d like to think that some com­pa­nies pick up the slack because it’s a good thing. And it’s the cheap­est invest­ment in stan­dards you’ll get. If not, then we’ll have anoth­er talk and we’ll see how it can go. I don’t know if that’s a real prob­lem yet.

Malamud: You talk about the increase in size of the IETF. Can we do use­ful work in work­ing groups in a assem­bly of a thou­sand people?

Deutsch: Interesting. There was a debate a cou­ple of—just bef—it was­n’t the last— I guess it was about six months ago; let’s sort of nail down a time here. About all this incred­i­ble rise, and what’re we gonna do, do we start train­ing cours­es. Somebody sug­gest­ed an entrance exam. Quite frankly if there’d been an entrance exam I nev­er would’ve start­ed going. And I like to think that I can con­tribute to the process.

So I did a post­ing at one point to the IETF mail­ing list and said, is there any­one out there who can hon­est­ly claim that their work­ing group lost a sig­nif­i­cant amount of time because of the large num­ber of peo­ple in the room who weren’t up to speed. No one answered. Now, that maybe nobody ever answers my post­ings, but I think more appro­pri­ate is to conclude…well, no they’re actu­al­ly not get­ting in the way. It is a fair­ly cheap way to train your peo­ple. Send them to the IETF and they find out what’s going on.

On the oth­er hand, you’ll get the inter­est­ing idea from the strangest per­son. You don’t know if this guy out from left field hap­pens to have just the con­nec­tion, just the answer we need to the prob­lem. I would encour­age… Now, I don’t want 30,000 peo­ple there. I go to Interop for dif­fer­ent rea­sons, but it’s a com­plete­ly dif­fer­ent expe­ri­ence giv­en the scale. But cur­rent­ly I think work is get­ting done. Because the inter­est­ed peo­ple sit down in the front. They par­tic­i­pate. Most work is done by a small core group, and the oth­ers act as sort of a san­i­ty check. And you can have a fair­ly large num­ber of peo­ple rep­re­sent­ing the user com­mu­ni­ty doing san­i­ty checks with­out it real­ly inter­fer­ing. And I think right now, cer­tain­ly in the areas I work in, which I don’t try to do rout­ing and I don’t try to do very low-level stuff. But the areas I work in, I can claim right now that we’re get­ting work done. And the large num­bers of peo­ple aren’t par­tic­u­lar­ly get­ting in the way. And we’re get­ting more work done, you know. There’s a lot of work to do. I would­n’t dis­cour­age any vol­un­teer that wants to come and help. So, at the moment I think it’s a man­age­able problem.

Malamud: You start­ed a com­pa­ny, and you did­n’t call it WAIS Inc., or Archie Inc., you called it Bunyip Information Systems. 

Deutsch: Yeah. Why not Archie—

Malamud: What is Bunyip?

Deutsch: Alright. This is a great ques­tion, and which I think means from the mar­ket­ing point of view it’s a great com­pa­ny name since you have to ask. I have rather check­ered his­to­ry. I was born in California, I grew up for a long time in Australia, now I live in Canada, and I moved around a lot. One point in my mis­spent youth, I read a book, in Australia. One of the lead char­ac­ters was Bunyip MacDuff. Now I don’t remem­ber too much about the book but I remem­bered the name and thought this is a nice name, has ties to a Australia because that’s where I read the book. And it just seemed like it’s a kind of name that peo­ple would remem­ber. It’s also the ​“IP” in the title, ​“Bunny IP” as some peo­ple call it, which does­n’t hurt. And it just…it’s some­thing that was always down in the sludge of my sub­con­scious that seemed to be a nice name.

Now, when we went to name the com­pa­ny, we— One obvi­ous choice was to try to get Archie in the title but…we’re not actu­al­ly in the soft­ware busi­ness, and we’re not in the Archie busi­ness. We like to think we’re the pub­lish­ing busi­ness. We make infor­ma­tion avail­able to peo­ple, we process it, we take it and do some value-added to it, and give it back to you. And so real­ly we want­ed some­thing that was divorced from an actu­al ser­vice or an actu­al pro­to­col. Because I real­ly think the impor­tant thing to real­ize is, nobody cares. The aver­age user does­n’t care what the pro­to­col is. They want the information. 

And I like to think that we’re try­ing to take a very broad, very gen­er­al, very non-protocol-specific approach. If you’ve got an access method that will get me use­ful col­lec­tions of infor­ma­tion, I will try to write an inter­face to it, or find a gate­way that does get to you. And I think that’s a very very Internet way of doing things. You’ve got use­ful stuff, I’ll do the hack that gets me to it. Somebody else gets use­ful stuff, I’ll do anoth­er hack. Now, if we do it right, it’s not real­ly hack­ing, its component-by-component piece­wise build­ing. I mean that’s how you build brick walls, one brick at a time.

So yeah, basi­cal­ly that’s what it is. It was an attempt to step back from a spe­cial ser­vice, one pro­to­col, and say you know, we’re not so [flip­pant?]. We can remem­ber our child­hoods. We don’t remem­ber all the details. But…yeah, we’re not so bad. That’s sort of where it came.

Malamud: Well, there you have it. We’ve been talk­ing to Peter Deutsch and this has been Geek of the Week. 

Malamud: This has been Geek of the Week, brought to you by Sun Microsystems, and by O’Reilly & Associates. To pur­chase an audio cas­sette or audio CD of this pro­gram, send elec­tron­ic mail to radio@​ora.​com.

Internet Talk Radio, the medi­um is the message.