I have a work unit being processed that is overdue. If I check for messages using the manager, I see, as you might expect, a warning that I might not get credit for the work unit and I should consider aborting it.

I let it run because I wonder if it is overdue on other computers, as well.

I suggest a feature for checking overdue units, and allowing the server to request these be aborted:

* When a work unit is overdue, the client periodically reports information about progress on the work unit to the server.
* When the server has received, from some computers, completed results for work unit that are overdue on other computers, it records an "abort request pending" for those that are overdue.
* When the clients "check in", the server can send abort requests for appropriate work units.

As far as I know, this is already added in the new upcoming BOINC 6.6 clients. Which hopefully will be released next week.

---

This probably won't work nearly as well as people think it will work due to the simple fact that in many, many cases the client won't receive the abort order until it's too late. Why? Because they'll be in communications deferral for up to 24 hours. Perhaps 6.6 clients will contact projects for which they have overdue tasks more frequently, for example every 15 minutes?

A 24 hour delay would work in my case. The estimated time remaining on a work unit already 2 days overdue is 180 hours. I don't want to abort it, because I keep imagining other people given the same work unit have a similar estimate.

A 24 hour delay would work in my case. The estimated time remaining on a work unit already 2 days overdue is 180 hours. I don't want to abort it, because I keep imagining other people given the same work unit have a similar estimate.

Then that feature already exists, I think since before 6.2. Server can send an abort request to the client.

It's especially useful when the WU isn't overdue, but it's not needed anymore, because other replicas ("results") from other computers already arrived.

---