Interview with Carsten Gundlach

0:00 All set. So it's the 24th of February, 2000, at noon, and I'm being with Carson Dunlap. So, well, we were just commenting on the European Network proposal. So I gather that it's been, because I remember when I was in Southampton three years ago now, which is the last time I was here, they were just, I think, embarking on a proposal at that time for the same idea, generalizing. And since then I gather it's been rejected twice, but now it's actually been accepted. Yes, well I can't really tell you anything about that because obviously I wasn't involved then. That's right. And even now I'm not officially involved. I just find it curious that I happen to arrive back at just the time when it's being discussed again, because now it's arrived. And so the issue at the moment is that, having been accepted, there are a number of post-docs available to be shared out amongst the network. and actually at the moment you're in the process of looking at candidates for that. Now, would your research be in any way connected with what Southampton is proposing to do under this network, or would it be quite separate? Well, probably only on the margins. Well, the research proposed by the network I think is a bit of a catch-all. Basically, the money was first, and then the idea of getting some of that money and therefore finding a common topic for several institutions. It doesn't mean that the research isn't serious, but I think perhaps these several institutions wouldn't have come together in a network if it wasn't for the money. So I think that's why I'm slightly not sure about the merit of this thing as a way of funding research. So, in Southampton, I believe the proposal is to do two fairly separate but related things. One is to introduce Cauchy characteristic matching to a three-dimensional cactus code so that you'd have proper outer boundary conditions in a non-linear evolution. So that's one project. The other project would be to work on perturbations of neutron stars, which, you know, the common theme is sources of gravitational waves, but that's about as far as it goes. They're really two quite separate projects, and I think what the poster itself would do

2:30 would depend on who we can get, basically. And perhaps also a bit on politics within the group and the collaboration. So I'd probably be sort of involved in the sidelines of both, either thing, but not in the center. Obviously the central person for CCM would be Ray, and for the Neutral Star project would be Niels. I was just curious when you say that in a certain sense that the money was there for a network of this type, do you mean that there's money available for large networks like this or that there was some sense that there would be money available for specifically a gravitational wave project? No, I think the European Union announces that it has so many millions to spend on funding research networks. and then people thought they might try to divert something of this to their own research. And it may turn out good, but in the end it's just not the most controlled way of funding research. So I think the funding system where you fund individual projects is more efficient. And as you said, Niels was saying, there is a parallel with a grand challenge for calculating very similar things again sources of gravitational waves that case was specifically the binary black hole problem whereas the European network is even more ambitious in a way, not more wider Right, the scope is even wider Right and again I think there in the States the motivation may have been the availability of funding rather than intrinsic need to bring collaboration together and I have a feeling that there now the grand challenge is dead and has totally failed its objectives. People are relieved that they can do what they think is needs doing rather than trying to stick to a sort of common program that nobody quite believed in. And I hope that's not going to happen here. I guess there is a certain similarity because I guess in the US case also there was a big government program for these type of big collaborations that had been set up. So people were looking at that big part of money. One thing that I was curious about, I don't know if this is at all intentional in the European program, but one thing that struck me looking at the history of GRR from the period of the 50s and 60s, as I did in a previous line of research, was that you could sort of see how groups seemed to cross-fertilize each other, or at least establish collaborations, simply by exchanging graduate students and postdocs.

5:00 So I was wondering if the idea of moving postdocs around Europe, which seems to be the essential aim of the PMR and network-style program, That's the clear name of the program. It just seems we have very many European postdocs you want to send around. Right. So does. Well, I had a postdoc. We've just hired somebody for, and I think there were equal numbers of European, Indian, and American candidates Of course, if it was a position in the States, the percentage of American candidates applying would be even larger, I guess. But for European position, European candidates were only about a third. So obviously, once you restrict yourself to European candidates, you've already got to eliminate two-thirds of the good candidates, if you assume they come in all countries, which they do. So, I think, for funding science, it's not such a good thing as it may be for getting Europe together. So there's a political aspect to it that is strong. As a matter of fact, if it was going to help science, probably the best thing it could do would be moving people from Europe to the US or vice versa. Well, there's NATO scholarships. I'm not sure those still existed, but they did. They do, I think. Yeah, and those were specifically moving NATO, and that usually meant going between Europe and the States. They also for political reasons, but accidentally it might be more useful to science. Right. Another thing I was curious about is if, since you've been at Potsdam and now you're here, if in some way the shape of the way in this subject, general relativities, I suppose. numerical relativity and astrophysical sources of gravitational waves in particular, becomes a little bit like, in Europe, a little bit like there's one big center at Potsdam and the other groups are sort of satellites of it, just because that group is so large relative to the numbers.

7:30 Well, I think there was a feeling that that would happen for a while. My feeling is, that probably would be debatable, but my feeling is it hasn't happened. So the Potsdam group is still very large, but it hasn't, it's underperforming, I think. It's not pulling its weight. And I think that's, on the classical side, of course, now it's become more of a quantum group. They have now, well, this is in fact not strictly publicly, but they have now appointed a second effectively quantum director. So now I have two quantum directors and one classical. So that's going to change its direction. And perhaps none of the two quantum directors are really world class. So that's not a good thing. So, but on the classical side of the Potsdam Institute, perhaps there isn't really enough of a concentrated program. It's a large seems to work together very much as a group. I mean, that work happens, tends to happen in all large groups, I think. But numbers really don't make for a big program. So I think when it comes to numerical relativity, the area that I know best is, well, it's a three-center thing. There's Potsdam, there's three and a half, perhaps. It was Cornell, Penn State, to some extent, Champaign. When I say a center, this means one or two people, right? You don't really need big groups to do good work. Sure. So I don't believe in numbers to do good research. Right. So even though there obviously has been some sense in the American Relativity that there'd be an advantage to having big numbers of people to tackle big problems? For instance, people setting up things like the Grand Challenge. Do you think that probably it's still a case where a more traditional style of working with just a couple of people who are really doing the work as justice or more effective than a large ground branch? I think it depends on how it's organized. Well, you would know a lot more about LIGO than I do, obviously, but it seems if there's

10:00 a leader who's both accepted as a scientific leader and a good organizer, then things can work and then people can pull together. If that isn't the case, as was the case in the Grand Challenge or is the case in the European thing, then I think having a common framework that everybody's tied to because as they said, that's what they would do, is probably an impediment rather than help. Already we sort of say, we can't work on this because the other group is working on that and they have their little thick mark to indicate that they are the main group. So that's not actually helpful, I find. So it tends to constrain the way that people might naturally choose their research directions? Well, yes. there's that effect and I think it's more of a negative effect perhaps than a positive effect and is that likely to be a particular problem because I'm actually quite really quite negative about this I think not only is it a bad use of money I think this problem may actually be worse to science than having no problem at all not even saying that the return on the money is low may be negative well that's interesting so Is that partly because if the state of research at the moment is such that one doesn't yet know what are going to be the most fruitful avenues to take, for instance, in numerical relativity, so if you have a program lined up or a program sort of set out and a way of working and people are, to some extent, constrained in what they're going to do next, that you lose the flexibility that's necessary at this stage? And I was wondering if there's an additional factor besides the one you alluded to where you have, if you have a sufficiently able... I think the negative part is that one may feel obliged morally to work on something that one doesn't think is necessarily the most interesting thing at the time. I think research should be interest-driven. I think there's also a negative effect that people get disillusioned. I think that happened at the Grand Challenge. You know, it gave the field a bad name, obviously, that's a fact, and it also disillusioned people within the field. So I think those are negative effects. I'm not trying to judge it. The Grand Challenge may still have had a net positive effect overall just because we did some fun, some good research. But it clearly didn't work together as a concerted

12:30 effort, as it was supposed to be. So, for instance, to take an example between Southampton and Potsdam is a concrete one. If it's successful, for instance, would marrying the Cauchy characteristic matching technique that has been developed here already to the Cactus Code would be in itself a useful... Oh yeah, if it were, it would be extremely useful. It would be one of many useful things I could do. I would not rank it among the top three most important bottlenecks in numerical activity at the moment. Seeing as what one can get by for the moment with doing funny things at the outer boundary. But if it worked, then it would certainly be helpful. Perhaps then we find out it's more useful than we think nowadays. That could well happen. The big question is, will it work or not? There was an attempt to do exactly that with Grand Challenge, to combine the Grand Challenge 3 plus 1 code with a characteristic code in the Pittsburgh group. And they delivered the code as unstable, and they've given up effectively. So that could happen once again, I don't know. The problem is there's nobody there who's really willing to lead that. So CCM has been a pet thing of the self-haven group for 10 years. So obviously they put it down again, but there's perhaps not the feeling of overwhelming enthusiasm and optimism that this requires. And then if you get a postdoc, who perhaps wouldn't have picked that as his or her topic to do either make him or her do that just because it's on the list and I think it's not promising so again it might be better to just let that wait until somebody and either until either the problem becomes more urgent or somebody actually finds it more interesting so the advantage then of not having these big collaborations where

15:00 The disadvantage of that and the corresponding advantage of letting people look for their own funding individually is that at least if it's people getting their own money, they will hopefully be more committed to the particular research that's set up. I think that's the main difference between the two approaches. At least people work on something they find interesting. one doesn't have the illusion that we can at the same time solve all outstanding problems just giving them to some poster you have to sort of pick what is your most important priority working it yourself perhaps hire one poster to help you right sure, so it's one another way of looking at it, it's one thing to get the money for a large number of posters all at once but still And the research directions that would probably be most effectively undertaken by those postdocs who are going to do the work would probably be the ones that they themselves were very interested in. So if you were saying beforehand that they were going to get directed in a certain way just because of where they ended up in a big program, that might have been most effective. In the European Union thing, there's two things that are noticeable. You talked about this ritual exchange of postdocs, which is a way of propagating science, that's perfectly true. But that's not happening just because I think none of the candidates, or perhaps one of them, Robert, who is graduating here, actually come out of the network. So we have sent Dennis to, well, Dennis has gone to Potsdam last fall. this has nothing to do with the European Union and Robert will probably go somewhere within this within the European Union but nobody has come out of AI in the last year or two who actually we could even possibly hire so it doesn't seem to work all that well they don't have a lot of students there so that's one thing it actually isn't an exchange Robert seems to be getting people in from elsewhere Those tend not to be very good people. The other thing is that practically none of the candidates are really numerical people. This is almost exclusively a numerical project. There's some analytic work involved, but I think most effective at some stage or another is numerical. I have a numerically evolving perturbations

17:30 or numerically evolving the full non-linear equations. And very few of the applicants have numerical expertise. And that's something that we've discussed within this group. I can't speak for everybody else, but within this group, it seems that Niels and I agreed that we basically don't have suitable candidates, so that should make us think about the usefulness of this whole initiative. You could get them to the States if you paid them $50,000 in order to outbid the Americans, but the world market for people with numerical expertise is very, very small, I think there we can't really compete. Wow. The big bucks are anywhere. Well, I mean, big-ish. 50,000 is still not big. It's compared to... You know, there's, I think, good numerical postdocs probably have a choice of positions in the States at the moment. So is what's required in Europe actually more graduate students than postdocs in the sense that there needs to be more people trained up in numerical relativity? I think so. I think there's a shortage of trained people to work in numerical relativity. And I think the reason is that, again, I wish I knew how this related to other fields, but in American relativity there's a sort of joking tradition that, you know, as soon as you have a faculty position, you stop coding and let the postdocs do the work. I'm just seeing this happen to myself, obviously. But that's something that we joked about a lot at the Barbara American relativity meeting. So that That means, effectively, that not a lot of training of graduate students is taking place in American relativity, certainly not in Europe, so I think that may be part of the problem. So that's why the post-docs seem to be scarce. They always seem to sort of work on the edges where they do perturbative stuff or gravitational wave analysis, that kind of thing, but there's a bit of a shortage of students getting trained

20:00 in key numerical relativity techniques. Another reason why that, in my opinion, is happening is perhaps that it's a double challenge. You have to both learn the numerics and you have to learn GR. When we last talked, in fact, in Potsdam, I think I mentioned that aspect in my mind then a lot, that we had a few students who never learned the GR because the computing took over quite naturally, and those people then very naturally gravitate towards non-academic jobs afterwards and are also lost to the system. I do actually remember you mentioning that at the time and it struck me as interesting so it's still probably a problem that's there I think that's not happening with our students in this group at the moment but I think it may be happening elsewhere in Potsdam my impression also was that they didn't have very many students relative to the postdocs and I think there it's the case that all the work is done by postdocs The only person who could have students there is Ed, I guess, in the American group, and he's only one person compared to quite a big group, so they're not turning out as many postdocs as one would expect from such a big group. Sorry, I didn't mean postdocs, I meant graduate students. So, part of the problem, as you were saying, is that the postdocs do all the work and then, for various reasons, when they become tenured, as it were, they then stop doing the hardcore numerical work. Well, I don't know. That's based on two small numbers. Well, I think in this group, Ray doesn't do any coding. I haven't done any coding since I've arrived here, even though I still hope I will one day. Ed doesn't do any coding in Potsdam. I'm trying to think of other groups. Stu Shapiro doesn't do any coding.

22:30 I think it's a tendency that the older people don't do much coding. It may be just because to do that you need large chunks of coherent time, I think it's It's easier to do pen and paper work, pen and paper research in small intervals than coding. You need to concentrate on it. Yeah, I think so. If you have lecturing and sort of things to do, then it may be easier to fit in analytical research than maybe an explanation or speculating. So, when you do find a time to get back to real numerical relativity work, what kind of issues do you see interesting you, since you get back to the topic of interest during research? Well, there's two things I think are the key problems. One is black hole excision. I think actually most people would agree with me, that's not very controversial. of black hole excision, so dealing with black holes in nonlinear evolutions, and the other one is probably coordinates. As soon as you have two black holes going around, and not just a single black hole or two black holes effectively head-on simulation, we'll have troubles with coordinates. And again, there's a lot of theoretical work about, let's say theoretical propositions about what coordinate systems to use. There's this one by Brady, Crayton, and Thorne that has been quite popular. There's the old maximal slicing, minimal distortion. Smart York comes back 20 or 30 years. David Garfinkel and I have been pushing another one. But really, somebody needs to go and try these things out. And that's only just starting because vector elliptic solvers with holes, the holes being the black holes, are only just becoming available. So again, that's sort of really driven by numerical technology. And there's one big exception that people who have permanent positions don't code, that is Bernd Bruggemann, who actually works on these vector elliptic solvers. So he's sort of one key person in the sense that nobody else wants to work with this horrible numerical stuff.

25:00 And the other thing is black hole excision, where many people are working on it and nobody has anything great to show. There's the effectively grand challenge stuff that is now carried on by Louis Lehner, where you sort of do a brute force approach, where you do finite differencing and do something funny at the boundaries, and that can be kept stable for a finite time and then blows up. And that's also the approach I intend to go back to for want of something better. And there are approaches that in principle are elegant, but really haven't worked yet either. For example, you could have Cauchy characteristic matching, but now do null coordinates that go into the black hole. It's a different thing. You'd have a region around the black hole where you have null coordinates that you match just outside the black hole, which would, I think, be the ideal solution if Cauchy-Grasher-List-imatching worked. So that's another reason I'd forgotten earlier for wanting that to work. Well, there is the approach of excising a cube that the Kohl-El group has started, which is conceptually extremely ugly, but they have had some success with it. The question is if it will work for things like two black holes. So I think that's a topic where everybody agrees it's perhaps the one most important problem in the binary black hole problem, but nobody has any sort of obviously wonderful answers. Right. The progress seems to be very slow. In the idea that you mentioned of extending the kosher characteristic matching, is the the idea there to, as it were, bring the characteristic area much closer to the source? Yeah, well, you'd have two characteristic regions. I mean, you could have kosher characteristic matching outside, and you march on a very, very large sphere where the gravitational field is already weak. But you could also match on a small sphere that just surrounds a black hole. And then you'd have, instead of outgoing null coordinates, you have ingoing null coordinates. So you won't have this when you take the . Oh, I know. Yeah, please do. But now I remember hearing about it before. So you'd have matching. If you had outgoing matching, you would have, so this is the LA radius. You'd have outgoing null surfaces.

27:30 And the idea is that you can actually compact define this outgoing null coordinates. In the interior region, you'd sort of have ingoing-out coordinates, and here would be the venturizer, and you just go inside the venturizer, and obviously you wouldn't need a boundary condition, because this is just a surface, so that's the idea. So that this surface where you match here, which is actually time-like, let's say R is equal to 3M or something like that, something just outside the black hole. And here is R is inside the black hole. This is actually the ventralizing. Well, of course, the point is you don't know where the ventralizing is, but you put the matching at the end of the calculation. And that will continue to work because you really don't meet any boundary conditions for no code. Your Cauchy region, in other words, would be completely isolated from the horizon today. Yeah, you wouldn't have anything special. You sort of try to have a space-like boundary here for Cauchy code that you have to do something special. Physically, again, you don't need a boundary condition. So that would be very elegant. Yeah, I was just going to say that would be an elegant solution. So obviously that would be one big motivation. Well, again, it's not obvious that it would be. So perhaps I was too optimistic. Because obviously there you have ingoing null rays. So they're focusing. And you know they will form caustics. So you can only hope that you can get far enough in to be safe numerically before caustics form. And that may be harder in a situation where the black hole is not even approximately spherical. So it would have to be tried, I think. Even though it's a nice application of CCM, again, a risky one, but for me, that's not a terribly strong reason to try CCM. Because CCM is itself so horrible and totally unstable. Well, sure. Obviously, if it had problems, then doing it twice might make it really worse. Well, you're trying to fix one instability with another one, in a way.

30:00 Something else I find in my research is that whenever I try to do numerical relativity, that is a binary black hole problem. My other research area, critical phenomena, sort of creeps up and takes up all available time. And in a way I think it's a less interesting area, less important to the future of research. It's sort of more like a neat little thing I can do well. Because I can do it well and because not many other people are doing it, it sort of naturally makes more claims on my time. So that's a particular problem I face. Right. Yeah, having the two different interests. So, I guess one quick question I should ask is, what do you think have been the main areas of progress in the American relativity since the last time we spoke? I think there's some progress in fluids in neutron stars there seems to be a little progress made by the neutron star grand challenge although they are over-advertised to a ridiculous extent so it's hard to see what they're really doing. But that seems to be getting somewhere. And I think also there's a lot of good work on neutron stars by the Meudon group using spectral methods. And that's focusing on single neutron stars and becoming very astrophysically realistic. Whereas what the Washington University group is doing is extremely unrealistic just because in 3D with two neutron stars on the you can't be realistic. But they seem to be getting there, so that's something. Another thing is, I think that people are starting to use coordinate systems with shifts as a matter of course, and that's getting somewhere. I think Louis Lehner is pushing the boundaries of the old grand challenge approach. I think that's always a good thing

32:30 to go be conservative and try to make the best out of what you have. And so, in a way, the state-of-the-art of the binary black hole problem is something very conservative, and that won't be the thing to do in the future, but it's the best we have at the moment. That's the simulations. There's no breakthroughs, I'd say. There haven't been any breakthroughs in the last few years. Is that another reason why Numerical relativity is a difficult area, as it were, because one, even if one is adopting an interest-led approach, there's always a certain inherent conservatism, because if you've developed a big code to work on a problem in a certain way, then you have a distinct investment in sticking to that method, even if it has encountered difficulty. Well, that's always true. I think it's independent of collaborations. Sure, that's what I mean, but that it's... It's terrible if you have to stick to somebody else's bestness. That's right. That's even worse. That's true. That's a good way of putting it. I mean, is that something that's different between numerical relativity and analytical work, or is it really just as true? I couldn't say. I think that's just true in research. So yes, there's some developments that are positive. I was still thinking of things. One is that Chactus IV, however good or bad it is, is now being quite widely adapted as a platform for parallelism, and perhaps in the near future for adaptive mesh refinement. And I think that's good, because it makes the investment you have smaller. So it makes people more flexible, and allows players like me to play. who sort of otherwise would have to start from scratch. So I think that's a positive development, although, again, it's been over-advertised, but that is something good. For example, the neutron star ground challenges go into Cactus 4 with the existing code, and the Penn State group has sort of something that they call Agava that you've probably heard about. It's based on the Cactus infrastructure, even though it's not exactly the same as what the Potsdam group used. And so I think that's a good thing. The other one is something I noticed at the Santa Barbara meeting,

35:00 is that people are becoming more aware about formulations of the Einstein equations that they use, and also the stability of numerical schemes. These are two different issues. They tended to be mixed up in people's minds, and I think that has been clarified. not in the mind of the sort of clear-thinking people, but in sort of the average mind, sort of communal mind of the community, I think people have... It's becoming clear that there's an infinite number of formulations to use, and none of them is obviously the best, so it's a matter of trying it out. And the other one is that... We often have been working with codes where we didn't know if the underlying numerical scheme was actually stable in the volume, there's always problems with boundary conditions, which, of course, these things can't, in a low-linear code, can't be decoupled. So I think people are paying more attention to stability problems. The other thing is hyperbolic formulations, which were sort of oversold by some people and now there was a backlash and now I think they're sort of coming back again and people are seeing that there really are advantages but there's a more cautious approach now so perhaps in the end we'll settle down to one or two formulations that are most useful although I wouldn't want to say which those are I'm interested actually in have become useful and widely used in the community. So you mentioned, for instance, the newest version of Cactus as an example of something that is becoming more widely used. If it was the newest version, that would be terrible. We hope it's a definite version. We, I mean, the users, are not really involved in writing it. But it's supposed to be no longer changing. And it is more user-friendly than Cactus 3, the version of it, with. So I think there will be a gradual move over to that and hopefully there will be a stable platform that will actually be used for a few years. So Cactus is fairly widely used? Well, it's used by several different groups, which is the only code that can be sent.

37:30 And are there other... Well, the Cornell group still don't use it and they made a conscious decision not to, But that's perhaps fair enough, because I think they'd rather start from scratch each time and keep control of their code and have a small code they can still read. Whereas, you know, if you work with CACPIS, you can't read your own code. So, there's different approaches. And you mentioned that the Cactus 4 is much more user-friendly, so does it work out in practice as being a relatively straightforward way of working that you can write your own bit of code that will then slot into and take advantage of the main Cactus code? Is it transparent, working on how you ruined it? Yeah, I think so. I noticed that sort of people at the edges work with Cactus. There's this post-doc applicant from India. So it works with Cactus. Somebody clearly has never been outside India, so you know, you can actually take this from the web and get it to work without having somebody in the office next to you to explain it to you. So that, I think, is a fairly strong test of working with somebody else's code. That is, yeah. How useful it is to do something totally new. You never know, because once you want to do something new, you'll realize that that kind of thing hasn't been foreseen in Cactus. On the other hand, I think it's so modular now that perhaps things can be done. So if you needed some totally different kind of grid or... I can't make this up because precisely I don't know what people would come up with next, but some kind of structure that doesn't fit into calculus perhaps it can be extended, perhaps it has been really made as modular as it can be So there at least is a reasonable amount of flexibility built into it Which also has made it totally unreadable as a piece of code, but the idea is that it's as modular as people knew how to make it at the time, of course, there's always something that you may not, some application

40:00 Is there other examples of tools that have been developed within the community that are becoming widely used? You mentioned, for instance, this work of . Oh, well, that's within Cactus. So I think that's, you know, so elliptic solvers, for example, are not exchanged normally. and they are something I find particularly unpleasant to work with so I'd be glad if some of those were made public in Cactus people tend to write their own so I think not a lot of tools are being exchanged and there is something to be said for that there is really an enormous advantage in starting from scratch and writing your own code and the sort of classical example for that in people's minds is David Garfinkel has written an axisymmetric vacuum gravity code axisymmetry was the cutting edge until very recently and still is something I would not think of writing from scratch because I think it's difficult but you know he's shown that previous effort you can actually do it very simply in about one year of his life with an incredible teaching load he's written this sort of thousand line code it's incredibly small and it works and so there is something to be said for stepping back and sort of take people's experience but not the actual code and starting from scratch if that can be done in 3D I don't know in axisymmetry I think it is being done I think there people really start from scratch with their new project, and that's perhaps a good thing. So, two things can happen. One is that actual tools, actual code can become more widely distributed and give people a certain level. But in addition, the actual experience that people have in developing these code becomes something of a communal property as well in some sense that people can then take advantage of when they're writing from scratch. Not as much as it could be. Something I think everybody agrees and that not enough negative results or technical tricks are being published. Yes And I think that will never change perhaps will change a little if people are more aware of the fact

42:30 But of course these things also make for totally unreadable papers so nobody wants to review even being a referee for these totally unreadable papers you know, just it should be there just in the hope that somebody finds that useful piece of information in there I'm not sure what needs what the best way of disseminating information is perhaps exchange your graduate students yeah, I was just going to say since as it were negative information and the learned learning from a mistaken sound doesn't get published by an Irish, presumably The people who are able to take advantage of it are the people who've been in the very active groups at some point, or maybe even in several groups. Yes. And of course, there are people who actively muddy the waters, mainly Ed Seidel and Waiwa Soon, who would never admit anything not working. So, you know, that's not a constructive attitude. Sure, well, I guess people are, and obviously... overstating it, but they would write out things in an overly positive tone, and that makes it harder to get the relevant information. I suppose it's a common thing to not want to publish your negative, your lecturing. Things that didn't work out, you want to publish your failure. So I suppose while the guy sitting in India, for instance, is able to take advantage of Obviously, he's not going to be able to take advantage of the other wealth of experience, say that a graphic will take advantage of this. Presumably, there's actually no way that this can get transmitted to any other. That's interesting. I remember actually talking to John Massol, and he was saying about cactus that he saw as a wafer. getting faculty status and immediately stopping no more code he seems to give up all of the academic life altogether rather than founding a company

45:00 and is this company related to numerical relativity what does it do well he's this is all hearsay so you should ask him the cactus people about that but he's apparently founded a cactus commercialization company in competition to the one that is being found at Potsdam, which sort of seems a bit of an exaggeration. But you should talk to the Potsdam people about that. I don't know. They're trying to visit there. I was just curious. I thought he had some idea about distributing cactus. I remember him talking about it and saying that it would be a way for the third world to strike back in numerical relativity against them. first world, because you'd have access to code that would allow you to work. Well, I think to some extent that's true. The fact that Cactus is an open source is, well, this is nothing new in the last few years, but it's a great thing. So I think overall it's a good thing. One thing I've maybe forgotten in 10 years, but at the moment, is that Tom Goodale is not going into the Potsdam company. He is one of the two main officers of Cactus, Cactus 4, with Gabrielle Allen. Cactus 4, of course, has been written from scratch since Cactus 3. So the ideas are the same. There's probably not a lot of code left that John and what's his name wrote? Paul Walker. Yes, Paul Walker. And I thought, naturally, he'd sort of give up his PhD and go to that company, but he's hanging in there, so I think that's good, because that means that somebody will, one of the authors will hang around in academia. So that's a good thing. But to some extent, once the quote has been released, and it has been released, that's it. And nobody can go and spoil it after that. That's a wonderful thing about public sources. Yeah, sure. Yeah, it's interesting. And I've been interested to see, to get an impression of how widely used it is,

47:30 I think from a purely sociological point of view, CACTUS is an incredibly interesting phenomenon. It started as one more relative to the code, right, we had the Bono-Massau equations hard wired into it, and it still did when I joined it. They're completely out now. So it started as one thing and it ended up as a completely different thing and it's been through about three different generations of authors as well. Yeah, it's interesting. It's funny how this thing sort of took on a life of its own. Yeah, as you say, it is an interesting sociological phenomenon because it has a life of its own in some sense that even changes the direction and different people being in charge of it, it nevertheless seems to be some undercurrent that keeps it going. the good contributions of Potsdam to the natural relativity scene. Having oodles of money, they sort of spend some of it on cactus. That amount of pure computer science development in a relativity group would not have been possible elsewhere. That may be something to do. That might be part of the thing with LIGO too, but it's part of the various management crises that it's been successful, because as long as there's the oodles of money there, you can at least keep people there, and get new people in? I have no insight into LIGO at all, I'm afraid. Yeah, I'm just thinking about that. The only sort of, outside, the only thing I can see is that they're clearly leaders, everybody agrees, are good people in charge, to be in charge, right? Even in the American relativity community, Greece, we usually sort of say, what would Kip say about this, you know? That's a fairly, you know, given that this can't be forced, you know, So it's not related to either money, he has all his position. That's quite a good, quite impressive. Sure, there aren't many that have that stature. Well, that's very interesting. Carson, thanks very much. I guess I won't keep you any longer. It's probably not much time.

50:00 I think we go with about one. yeah I'm going to go to he's running it back on on I guess you missed something there most of it was technical talks but there was a lot of group discussions and I think there was a pent up need for open discussion and I think the relevant people weren't there so there was actually free discussion for example Matsner wasn't there most of the conference obviously he is justly or unjustly he is a big scapegoat of the grand challenge that would have been quite hard to have a discussion in his presence I'm not judging that so there was this discussion are we ideas limited or do we really have a problem that can be broken into parts I think the overwhelming feeling was it probably can't be broken up it was too early to break it up during the grand challenge on our papers, even over-cautious, and they say, well, we don't know at all what we're doing, and let's just be very modest and let everybody sort of go back to their small groups and then work away on it until we have something. I added the suggestion of having a collaborative review paper, something along the lines of living reviews, more collaborative effort, perhaps less definite, just sort of exchange the state-of-the-art results. I think about half the group was in favor and half was against, which obviously killed the project because you would have to have an overwhelming majority in favor for that to work. But that made me think that people really think that, I think they're so embarrassed about the grand challenge that they're over-cautious now. Perhaps they are just cautious, but it's not over-cautious. Everybody sort of has this disclaimer now saying, well, we don't have any good ideas, and we must have new ideas. Right. That's interesting. Yeah, I would have... I think there has been this backlash. So I was interested in Cactus, as it seemed to me, an example of an effort to meet this problem in a technological way, as it were, by writing code

52:30 so as to make a certain amount of the technology or tools that have been developed available to a wide number of people in a fairly independent way. I think it's not mainly a technological problem. There's no technological answer. What Cactus does effectively is give access to parallel computing to people who don't want to know about the detail of parallel computing. I think that's a big thing already, but perhaps not as much as has been advertised. I think when it comes to the relativity modules, they're perhaps not as widely used as others, and people might prefer to have proof-of-principle codes that are not parallel, or they're parallelized in some crude way of thumb. But there's synergy. You can still say, well, I have this proof-of-principle thing, And then when it works, and I really need parallelism, then I can put it into a cactus. And you know you can always do that later because the cactus won't go away. And you know that it can be done. So that's very good. I'm not sure it helps much with the actual collaboration on the problems that really limit us, and those problems are not problems of parallelism. I mean, if we had computers a thousand times as big as we have now, it would be hardly any better off than we are now. The idea is limited. say, you know, 10 to the 9 times bigger than we have them now to obviously unrealistic. But if you could go with a factor of 10 to the 3 in each dimension, then perhaps you could sort of try a brute force approach, but not really in the foreseeable future. Well, my impression talking to Ed a couple of years ago was that part of the idea with Cactus was that it would provide a way for people to work together collaboratively without having to coordinate their approaches that much, because if they wrote their codes so that they both slotted into Cactus, then their codes would sort of automatically work together. I don't know why that's not really the case. It should work, shouldn't it? It doesn't seem to be happening. I think it works better for analysis routines and for things that are sort of crucial. For example, when you start excising black holes and you have to, again, almost start from scratch and then you realize that putting it into cactus from the start

55:00 is actually much harder than if you had a simple code. Whereas if you have an analysis routine. That sort of is trivially linked to the rest of the code. It just sort of reads things. And it may be okay. It may be a good thing to put it there. Okay. So it's actually a useful tool for individuals because it saves them certain amounts of labor and they can take advantage of parallelization without having to become experts in that topic, for instance. But in fact, it probably doesn't seem to have much impact as far as being a common platform that everybody uses so they could actually all? Well, it's not a common platform everybody uses. It's a common platform two or three groups use, which is perhaps better than nothing. And why it's not being used by everybody, I don't know. I think it's just that some people want to keep control, not in a political sense, but just in an intellectual sense. They want to know what every part of the code is doing. I think it is a platform that has been recognized as a good way of making an existing code parallel after it's been written. That's somewhat different from this collaborative infrastructure thing that you mentioned. Yeah, I'm interested in the extent in which those initial ideals have been realized. I would be curious why other people think that Cactus isn't. Well, if I think, first of all, that canvas isn't being used as a collaborative infrastructure, why? I've been intending to go around and ask people. In fact, one of the things is that I started out a couple of years ago and I really went around and taught to all of the numerical groups as much as possible because it was an area I wasn't at all familiar with before. And since then, I've ended up sort of going back a bit and looking at the data analysis people and the astrophysics people. and in a way I'm only just getting back that took up so much time that I didn't miss things for the Santa Barbara meeting and so on so anyway I'm trying to pick up the story again I volunteered to give a sort of summary of that here and then because Ray was on sabbatical and we sort of haven't had group meetings for that reason this never happened but I could, if you have some more time in the afternoon and I could sort of go for my notes and just give you a mini-summary. Okay, that would be great. Quite helpful. Yes, to me, myself. Yeah, that would actually be great. I have tons of notes that we're just forgetting things I've heard.

57:30 Sure, okay. I'd like that, actually, so maybe we can do that after lunch. Okay. Okay, well, I might as well actually turn it off at this point. Yeah.