Interfaces and downtime

A message that arrives changes nothing by arriving

Messages are stored, acknowledged and put on a queue. What one means here is a decision somebody makes on the surface it belongs to, with that surface's permission, and then says so on the message.

The same message twice is still one message

Senders resend when an acknowledgement goes missing, which is what they should do. A resend is answered with the acknowledgement already given, and nothing happens a second time.

Nothing is thrown away

A message of a type a feed is not set up for is kept and held back with the reason, not dropped. A message nobody will act on is left with a reason that stays on the record.

One key delivers to one feed

A feed that receives HL7 names the one API key its sender delivers with, and a message sent with any other key is refused. A live feed needs its key, and a key held by one feed cannot be bound to another: give each sender its own.

No credential is kept here

A feed names where the other side's credential lives. Pasting the credential itself is refused, because a place to hide a secret in plain sight is worse than an empty field.

Nothing written during an outage goes into a record

A late entry is a note that something still has to be entered where it belongs. Enter it on that surface, then mark the note as caught up and name what you entered.

Both times are kept and only one is yours

You say when care happened. When it was written down here is recorded for you. The gap between the two is exactly what somebody reviewing the outage is reading.

An outage cannot be closed with work outstanding

Closing is refused while a late entry is still to be caught up. That is the failure this exists to prevent: the record says it was dealt with and the ward still has a pile of paper.

Set up a feed

1. Stay on Feeds and select Add a feed.
2. Give it a code and a name, and choose what it speaks and which way messages flow.
3. List the message types you have agreed with the other side. Anything else is kept and held back.
4. For a feed that receives HL7, choose its delivery key: an active key granted Deliver interface messages in Developer settings.
5. Put it live when both sides are ready. Hold it, saying why, if you need to stop accepting without losing the setup.

Work the queue

1. Open Waiting to see what has arrived and not been dealt with.
2. Open a message to read it and see what has already been tried.
3. Enter the record where it belongs, come back, and record what you did with its name.
4. Choose Entered and name the record, or Could not and say what stopped you. Nothing is chosen for you, and the attempt stays on the record.

Run an outage

1. Open Downtime and declare one, saying what is not available.
2. Write down what happens while it is down, with the time each thing happened.
3. Select It is back when the system returns.
4. Still to catch up lists what is left, oldest care first. Catch each entry up on the surface it belongs to, mark it caught up, and close the outage when nothing is left.