The BRIDGE Scenarios

Simple Scenario B: Sally Waits for Harry

• 2002.10.25 Initial posting of scenario. (cmr)
• 2002.10.28 Adding notes. (pli)
• 2002.10.30 Posted responeses (more questions, really.) (cmr)

---

CMR (2002.10.25): Ok, so what can we detect with respect to Harry? It would be nice to tell when he is getting ready to edit the abstract, but maybe we can only tell when he is in the full editor, or just if he is logged in (that would be less than minimal.) Let's put something in here we can achieve NOW and make it harder later.

PLI (2002.10.28): This gets us back to the "proximity" question. Abstractly, we could think of each object having a set of proximal objects, each of which having:

• A distance from the object
• A direction from or (more generally) relationship to the object
• A specification of privacy: how much should people looking at the proximal object know about exactly what I'm looking at

Our current scheme has two distances (logged on and at the same object/logged on but not looking at the same object) and strict privacy control. I think we can (and should) come up with a list of ways that such proximal object sets might be defined, but a default might be to define it based on position in the underlying hierarchical data organization:

• All objects n levels up or down could be informed
• Direction/relationship would be up, down, or adjacent
• Privacy could be based on direction, e.g., lower nodes would be told exactly what you were looking at, which upper nodes might just be informed of the folder/site you were in.

In practice we would likely only inform folder/site objects, just to save bandwidth.

CMR (2002.10.30): This proximity question is vital to what we are doing, but I think it shows up more in the scenario A than in this one. The issue I was trying to raise is what can the proxy see (I was not clear about that.) The proxy will likely be deployed to monitor just one object (or a set, but that would still put specific bounds on the proxy's horizon.) I think my question more relates to what an object is. Is a web page and object or is the full editor an object? Both? The full editor operates on the web page, so what does this relationship mean? Bottom lines though, can the proxy detect a user entering the full editor? Can the proxy detect a user editing a specific page? Any limits to any of this?

---

CMR (2002.10.25): How will she find out the proxy has completed the mission? Does the alert need to show up within the Java environment or can she be anywhere on her computer? Help me write the alert part in a way we can get to work quickly.

PLI (2002.10.28): I think the key here is to design the simplest possible abstractions for missions and alerts, so that deploying a proxy is just a matter of establishing a (mission, alert) pair such that something server-side can find it. When the mission says it has completed, the alert is invoked as passed information that the mission provides regarding its state at completion. One such concrete alert implementation might be a list of other alerts, in the order in which they should be tried: send alert to client and wait for response; if that fails attempt phone call; if that fails, send email. A next step in the design would be to enumerate enough of these to have a reasonable idea of what the interfaces (in the sense of API interfaces) for missions and alerts need to look like.

CMR (2002.10.30): That should work. But I still wonder if there are limits. I'm guessing that if someone is in the Java client we can pretty much make any notification happen at any time. However, if the proxy owner is in the web browser, what can we do? Users will think that being in a web browser looking at the site qualifies for the same level of service as being engaged through the Java client, but I'm not sure we can do that. I need some definition on the limits here so I can draw the scenario to some conclusion.

---

CMR (2002.10.25): The scenario ends when Sally and Harry connect somehow in the same place.

---