142
2. Minimal interaction time.As in the case of email, there is frequently a minimal pe-
riod of time that should elapse between activities on a channel. For an interactive
socket connection this period of time might be instantaneous, while for an email
connection, it might be a day or longer.
3. Autonomy.Some connections spontaneously produce new events, while others pro-
duce them on request. Autonomous connections have their own timetable for pro-
ducing events. For example, most laptops only have network connections available
when the user has chosen to take action.
4. Persistence.Some connections guarantee that information sent on them will persist.
This is true for a connection that creates a file containing the information sent to it
and is often true (or supposed to be) for remote server machines.
5. Interrupting.Some connections produce changes on their own, and others need to
be polled. Interrupting connections produce events on their own.
6. Timed.Some connections periodically produce new changes, or transmit changes
that have been accepted but not transmitted. Timed connections take action on their
own at specific intervals.
7. Pullable.Some connections take requests for particular change IDs which they will
attempt to produce at a later time. Pullable connections, when available, allow con-
flict resolution policies to automatically request missing changes.
This set of properties is exemplary, not exhaustive. Only a few of these properties, pul-
lability and directionality, affect the interface of a connection in terms of it available opera-
tions, as opposed to its behavior. The others only affect the way that connection will inter-
act with a given policy. Consider the pushy distribution policy described in the previous
section. Applied to a direct network connection it would provide services appropriate to an
interactive editing application. The same policy communicating with 3 collaborators over
email channels with 1-day transmission timers would maintain rough synchronization be-
2. Minimal interaction time.As in the case of email, there is frequently a minimal pe-
riod of time that should elapse between activities on a channel. For an interactive
socket connection this period of time might be instantaneous, while for an email
connection, it might be a day or longer.
3. Autonomy.Some connections spontaneously produce new events, while others pro-
duce them on request. Autonomous connections have their own timetable for pro-
ducing events. For example, most laptops only have network connections available
when the user has chosen to take action.
4. Persistence.Some connections guarantee that information sent on them will persist.
This is true for a connection that creates a file containing the information sent to it
and is often true (or supposed to be) for remote server machines.
5. Interrupting.Some connections produce changes on their own, and others need to
be polled. Interrupting connections produce events on their own.
6. Timed.Some connections periodically produce new changes, or transmit changes
that have been accepted but not transmitted. Timed connections take action on their
own at specific intervals.
7. Pullable.Some connections take requests for particular change IDs which they will
attempt to produce at a later time. Pullable connections, when available, allow con-
flict resolution policies to automatically request missing changes.
This set of properties is exemplary, not exhaustive. Only a few of these properties, pul-
lability and directionality, affect the interface of a connection in terms of it available opera-
tions, as opposed to its behavior. The others only affect the way that connection will inter-
act with a given policy. Consider the pushy distribution policy described in the previous
section. Applied to a direct network connection it would provide services appropriate to an
interactive editing application. The same policy communicating with 3 collaborators over
email channels with 1-day transmission timers would maintain rough synchronization be-
