The basic idea is to have a central computer (or computers) which accepts neutrino burst candidate messages from neutrino detectors around the world and sends an alarm message to astronomers (via astronomical bulletin or other paths) if it finds a coincidence within a short period of time (~tens of seconds). Experiments which have expressed interest are SuperK, MACRO, SNO, LVD and Baksan; others may also join (and in the future we may be able to include gravitational wave detectors such as LIGO, etc.) The coincidence will be "blind", i.e. will be done without polling other experiments if one message is received.

For a 2-fold coincidence, the rate of alarm messages from each experiment must not exceed ~1/month to have a low probability for a false alarm to be generated over a ~10 year time period.

At the neutrino conference in Helsinki in 1996, it was decided to set up a test computer for this system at SuperK. This has now been done; some prototype code is now running on an old Sun computer there and is ready for tests between different experiments. We'd like to get some more inter-experiment tests going soon.

The prototype code is in the form of fairly straightforward client/server code. It is written in C but is callable from Fortran. It has been tested on a number of platforms and should be very portable. The server code runs on the central machine (currently at superk). Client code runs on the individual experiments' machines; it sends datagrams via direct socket connection to the central machine (retrying if the connection fails). Alarm messages contain information about the UT time of the alarm, the experiment, and "significance" of the alarm (it's not clear how this last parameter will be quantified yet). The server machine accepts the messages, keeping all from the last day or so, and checks for coincidences within a one minute time window, according to the time stamps on the messages. The time window may be tightened later. If the server code finds a coincidence, it sends email to interested parties (later it may send directly to beepers, etc.).

So far, we have done preliminary tests between the central computer and the MACRO and SuperK computers. (Only dummy data records have been sent so far; real data will not be sent until intra-collaboration "protocols" for releasing data have been finalized.) We are happy to make the client code available to anyone from other experiments wanting to try some tests.

In the future, the "central" computer location may change; there may be more than one server computer in fact.

At some point over the next few months, there should be another meeting between representatives of the different experiments to decide on more of the details of the data exchange.