The National Seminar for Theoretical Computer Science, NSTCS, has now been around for 8 years. It has travelled across the country; being hosted at Chennai (91, 97), Calcutta (92), Kharagpur (93), Kanpur (94), Bombay (95), Banasthali (96) and Bhubaneshwar (98). It has seen major changes in format: from 18 contributed presentations and three tutorials in 91, to three tutorials and nine invited talks in 96, to 29 contributed presentations in 97. These changes have come about due to an altered perception of what NSTCS is meant to achieve and how it can best benefit the Indian computing science community. And change is inevitable; the existence of NSTCS changed the dynamics within the Indian community, changing the very needs that NSTCS sought to fulfill.
As one who has attended five of the eight NSTCSs held so far and has been party to several discussions on ``Why NSTCS'', I have had a fairly ``close-up'' view of the changing opinions. At the business meeting at Bhubaneshwar, extremes of opinion were reached, with some people arguing for stopping NSTCS altogether (replacing it with different kinds of events) and others whole heartedly wishing its continuance. Against this backdrop, in the few months following NSTCS-98, I conducted an email discussion with several people on the desired future of NSTCS. Below, I first recount my perception of the NSTCS's in the past, and then present the post-NSTCS-98 opinions of 3 people. Hopefully these opinions will find their impact on the future course of activities charted out by IARCS.
Major inputs have come from P S Thiagarajan (SMI Chennai) in 1991, Somenath Biswas (IIT Kanpur) in 1994, Madhavan Mukund (SMI, Chennai), S Arun Kumar (IIT Delhi) and V Vinay (IISc, Bangalore) in 1996, Abhiram Ranade (IIT Bombay), R Ramanujam (IMSc, Chennai) and Ashok Subramaniam (IISc, Bangalore) in 1997, Jaikumar Radhakrishnan (TIFR, Mumbai) and Venkatesh Raman (IMSc, Chennai) in 1998. Others who have also participated in such discussions include Gautam Biswas (IIT Kharagpur), Sounaka Mishra (ISI, Calcutta), K Rangarajan (MCC, Chennai), Sanghamitra Mohanty (Utkal Univ, Bhubaneshwar) and most of my colleagues in Chennai.
I recall the first announcement of the first NSTCS in 1991. It was a Call-for-Papers, in typical Conference CFP format. However, it also explicitly stated that the Proceedings were not copyrighted. The philosophy behind this was that NSTCS was to be used as a stepping stone by researchers in the Indian community to hone their skills at writing good research papers. Several members of the IARCS Steering Committee (it was not called any such thing then) had been concerned about the sharp decline of FST&TCS papers written by Indian authors, and felt that one reason was an inadequate understanding within large parts of the Indian community of what constitutes a good research paper. They hoped that by having stringent refereeing at NSTCS, and by sending back to the authors detailed constructive criticisms regarding their contributions, the overall quality of Indian submissions to FST&TCS would improve. I remember that authors of papers accepted at NSTCS-91 were urged by the NSTCS Chair to consider submitting their paper to FST&TCS-91, and I believe of the 18 NSTCS-91 papers, 6 can be found in FST&TCS-91. A cause for some satisfaction, one might say.
Another place where NSTCS was expected to help was in provided exposure to different research areas. It was unfortunately true (still is?) that research groups in different parts of the country did not interact much except at FST&TCS, so graduate students often had very limited awareness of research topics other than those pursued within their own groups. To rectify this, NSTCS tutorials were proposed. They were to be at a level where a student one year into a graduate program can follow what is going on, work out some exercises, perhaps even get excited enough about some problems to decide to work on them, and so develop more than a nodding acquaintance with the field. As to how many success stories the NSTCS tutorials have had, I am not too sure. Perhaps I can count myself as one: NSTCS-91 had a tutorial on Complexity Theory, which I was mildly interested in then, and post-PhD I have worked only in Complexity! There's more: during my graduate studies, I did not have any serious exposure to areas other that Automata Theory and Algorithms, so the tutorials helped me to at least know what area is about what, if not work in those areas.
So that was the spirit in which NSTCS started. Where did it head out from there? Soon there came a paradigm shift: several people felt that it was important for Indian researchers to get a chance to speak about their work to an audience outside their immediate group. Of course this is important; the point I'm making is, they felt NSTCS could serve this need. Refereeing thus changed gradually from a stringent search for potential quality to an elimination of ill-formed and out-of-scope ideas. This required less material for study, and the Call-For-Papers, as it was called till 1997, began to invite short 4-page abstracts. Often the response was even a half-page abstract. By 1998, the refereeing process had been reduced to merely returning papers not in mainstream theory of computing, and the main job of the Program Committee appeared to be providing some hopefully useful feedback at the presentations, and fixing up the tutorials and invited talks.
Invited talks? When did those come in? Around 1995, I think. The change in refereeing, coupled with vigorous advertising of NSTCS, had had the effect of getting submissions from more and more far-flung places. For several people from these groups, even the tutorials were somewhat advanced, and the idea of invited talks was mooted. These talks were meant to give wide shallow coverage of research areas. They were a great hit with much of the audience, but fixing up the talks became the Programme Committee's nightmare: getting people to talk is easier than it sounds. Also, by this time the great divide had become explicit: there are researchers from ``elitist'' Institutions (IITs, IISc, TIFR, IMSc, SMI, ISI, ...), and there are non-elite researchers from the Universities and colleges. What was exciting for the former was too advanced to be useful for the latter, what was appreciated by the latter seemed boring for the former. Timings convenient for the former clashed with college teaching schedules, college vacations were when several researchers who could wanted to visit abroad. And increasingly, people from the former group just stopped attending NSTCS, unless they could be persuaded to give an invited talk!
How could NSTCS help both groups? The best help it could give was by just existing, by providing an annual forum where people from both groups could meet. Was this enough? And hard on the heels of this came a niggling doubt: are the elitist researchers, already seasoned in the international setting of competitive research, being condescending towards their less fortunate colleagues? An uncomfortable thought; better to push it away and continue. But it would not be pushed away, and the Business Meeting at NSTCS-98 openly concluded that there are two different types of events we need, both of which cannot be clubbed into NSTCS. We need a regular interaction forum other than the international high-pressure FST&TCS, we also need schools and workshops on a much larger scale than the tutorials of NSTCS. Clearly, it is time for all IARCS members to come forward and take some initiative; the well-being of the Indian TCS community is the well-being of each individual in it, and with sufficient will, none of the problems that have cropped up are insurmountable.
Below are some opinions generated during the post-NSTCS-98 email discussion, with specific suggestions on what kind of events can be organised in NSTCS and in addition to it. To date, NSTCS-99 has not yet taken concrete shape, and hopefully it is not yet too late for some of these suggestions to be taken up in 1999 itself.
Suggestion 1: Given that the emphasis of NSTCS is not on high-quality submissions, the announcement should not even call itself a call-for-papers, which is suggestive of a conference. The announcement should be along the lines of (1) announcing the next annual meet of the TCS community, (2) promising invited talks and tutorials in leading areas, and (3) mentioning that there will also be sessions for participants to discuss ongoing work, and those who are interested in presenting something here should send an n page (n=4?) abstract by a fixed date, some basic screening/selection will happen, submittors will be notified by another fixed date ... .
And the abstracts need not be carried in the Proceedings. The proceedings will be a very good reference if they contain all the tutorials and invited talks.
Suggestion 2: An alternative to point (3) above is to abolish contributed papers altogether, especially receiving them in advance. We can never be halfway about this. Once a paper/abstract is published, people will quote it, and so we should do a serious refereeing. If that can't be done, forget soliciting papers to start with. We could have a time slot of an hour or so each day or even every afternoon for participants to present ongoing work.
It is often mentioned that this will discourage college teachers from attending. I think this is a bogus excuse. I think there are enough avenues to send the paper to. (After all, NSTCS is less than 10 years old. What happened before that?) The message should get across that this is a forum to get FEEDBACK, and not to publish one's paper. The fact that this message is not gone is evident from the several no-shows (i.e authors don't care about feedback) and the fact that in some presentations, the speakers blatantly ignore the feedback. So I think the National Seminar should not encourage people to use this avenue to ``publish'' junk papers.
( Moderator's Note: Since 1991, the NSTCS Proceedings have been in the form of a non-copyrighted Technical Report. (The single exception was NSTCS-96.) They usually carry the text of tutorials and invited talks, and the abstracts of presentations. In 1998, the NSTCS Proceedings did not carry even the abstracts. The reason was that though refereeing had become non-existent, there were cases of people citing an NSTCS ``paper'' in a list of publications. This raised the ethical question (mentioned in Suggestion 2) of whether the Program Committee can allow such a possibility at all if it is not doing serious refereeing.)
Suggestion 1: Given that there seems to be a shortage of people willing to commit to invited talks, we could perhaps have one session in the format of Journal Club talks. In this session (which also must be fixed up in advance but may be easier to fix), people present recent papers they have read (not necessarily their own) which they found particularly nice - some elegant theorem, a nice characterization, a new formalization ... . I think it will be easier to get speakers for this kind of session, because it requires a little less work than preparing an invited talk as a survey.
( Moderator's Note: This suggestion was implemented at NSTCS-98; there were two excellent and well-received Journal Club talks.)
Suggestion 2: Each tutorial could be given by more than one person (see Suggestion 1 under Tutorials below), and some of these speakers could include even senior PhD students.
Suggestion 3: Until there is sufficient interest from the premier research institutes,
I think they should continue perhaps with the following change. Instead of one person talking for three to four hours, one person can organize two to three speakers covering various aspects of the theme. This way the audience get variety (and so less boring), the stress is less for the speaker, and this will also encourage a little more attendance and participation. At the same time, the audience get a full picture of one theme. The speakers need not be faculty members, it could include even senior PhD students (to take care of lack of speakers).
I am against converting the tutorials (or the national seminar itself) into schools. The schools can be in addition to the national seminar. I personally have immensely benefitted from the tutorials, and I think several others have. These talks/tutorials give exposure to various topics for those who are not all that seriously interested in the topic to do research, but interested enough to know what's going on. (I think the schools cater to the former kind of people). Also the tutorials give an indication of who works where, on what topic and what is current in that topic etc. This will be particularly useful for college students and teachers. There is no other forum where Indian community, at large, can meet leisurely and discuss research.
So for these reasons, I am all for continuing the tutorials (in a strengthened way) and the seminar itself. We could have simply two to three tutorials or theme sessions in each seminar.
From college/university teachers: I think wide publicity in all engineering colleges and university is needed. We are not yet systematic enough about this. I think Aditya Sastry did a wonderful publicity job and so it got huge attendance (even though it was also summer and in a remote place).
From serious researchers at IITs etc: Since keeping up-to-date in your own research is difficult nowadays, very few people care about what is going on in other areas. I think you can't do much about such people. But I think the message would trickle down eventually that this is a relaxed forum to learn a lot of things. Again making these people talk (tutorials etc) will increase some sort of participation. We could also try hosting it in a cool place (like Kodai) even if it is the same venue every year. Then people will mark it in their calendar.
I have described below a few things that I believe are important for research, starting from the time a person decides to take it up. There are things a research student should find out before he takes up research actively. He must know what the different areas of research are, and what kind of skill and training is needed for working in them. Based on these, once the researcher has chosen an area of research he needs support for his research. He should have some means of obtaining up-to-date knowledge of his area. He should know what problems are important in his area, and why they are are important. He needs some means of acquiring the skills that are needed for working on problems in his area. He needs access to at least one experienced researcher, whom he can approach for advice and help.
NSTCS does not provide any such environment; it was not intended to. The goal of NSTCS is to provide a forum for interaction and exchange of ideas between researchers from different parts of the country. But where are the researchers? NSTCS today serves researchers mainly from some five institutions (which, incidentally, are well connected by email). These researchers travel to a new place every year to interact with each other. NSTCS would be great idea if universities all over the country were teeming with enthusiastic researchers, eager to talk about their work and to find out about the work of others. In the conditions prevalent today, NSTCS achieves very little.
In order to establish a culture of high quality research in the country, researchers must have an environment that provides the necessary support. I have two suggestions for creating, or at least simulating, such an environment.
We need surveys of different areas of research in general terms. These should be aimed at someone who wants to quickly find out what the area aims to study and what are the questions that it seeks to answer.
How does one come up with such guides? I think in areas where there is some expertise, we should organize ourselves into interest groups and study the existing research thoroughly. We should then formulate a programme for research describing the important open problems in the area and the background training required for working on them. The survey should mention work done in the country in the past and who is actively working in the area. In particular, a part of the survey should present the area as the researchers in the country see it: the problems they are working on, the problems they would like to see solved, and where this fits in the whole area. If people feel that they have access to experienced researchers, more and more researchers will feel encouraged to take up research in that area.
It is natural that those involved in going through the literature and writing the survey will acquire considerable expertise and knowledge of the area. A few of them can put together a course, perhaps a month of lectures, two or three hours a day. They could, if necessary, break up the area into background training (the mathematics etc.), the foundations and recent research, and design courses for each. If the area is well chosen, the courses designed will have lasting value. To reach the expertise to different places, we don't need to organize an entire conference. A couple of researchers can travel and give the course. If in each institution that hosts such a course, even one faculty member decides to pick up the material and go through it thoroughly we would have created a permanent resource at that place. The cost and administrative work for this will be much less than that required for organizing an NSTCS and the returns will be much higher.
I believe, as I said above, that NSTCS is good only if there is a culture of high quality research in different parts of the country. So, we must now work towards creating such a culture. We need good surveys meant for beginning researchers. These surveys can be published in the Newsletter; on they other hand, talks based on them can also presented. We can use NSTCS for this, which will now be devoted to a general survey of these areas, so that students who attend them can quickly find out what is there in each area, and whether they should pursue it. But this in itself is not enough. We need a follow up. The survey articles and NSTCS will only help develop good taste and, to some extent, confidence, but this will be converted into research only if there is expertise and a programme for research; this will come only from the short courses.
Including survey articles in the Newsletter: The EATCS Bulletin carries excellent columns on 4 subareas of tcs. The IARCS newsletter can carry one column, on any topical area of interest to some group in India. With the newsletter as yet being merely biannual, we don't even need to generate articles very fast - surely this should be feasible. Or does the Newsletter Editor feel otherwise - do contributors have to be chased and cajoled into writing?
One model I really like in the world of our physics colleagues is the Theoretical physics seminar circuit (TPSC), where speakers, rather than listeners, travel. We've been talking about this on and off and saying we're not yet organized enough to do this. Now that IARCS seems to be getting a formal existence, the time has come to begin these activities. (For instance, as a formal registered body, we can apply for governmental funding to sustain such an activity. )
This still leaves the question of NSTCS---the above activities can be independent of NSTCS. This brings me to a point raised by Abhiram earlier and by Jaikumar above---is the intermingling of researchers from the elite institutions and the rest visible at all in NSTCS? If that is the primary goal, NSTCS is not doing well at all. On the other hand, as Venkatesh has argued, there is perhaps a need for such an event even for the "elitist researchers". A forum where we can meet in a relaxed manner. So why not take the original goals of NSTCS and split them into two components which appear to be orthogonal? One will give rise to an NSTCS-like event for the better-placed researchers --- a small affair, maybe like the annual Dagstuhl school in Germany. The other will lead to courses and schools.
I think right now NSTCS is one event which tries to achieve almost everything that FST&TCS cannot. Unfortunately, co-FST&TCS is too large a set to be captured this way. I am all for breaking it up as above, keeping something good for everyone in different events. More events will need more voluntary participation, and there is tremendous inertia, even amongst the younger scientists, regarding attending an NSTCS-like event. But if one event has a higher payoff (the Dagstuhl-like event), I believe there will be more volunteers for the remaining events as well.