Comment 4 for bug 1402109

I have concerns about this patch that I discussed with Jason and Newell today. Fundamentally, I don't think we should be doing any de-duplication (or folding) of events being written to the database; the issue this bug presents is a display issue, not a performance issue.

We can fold on the output side; by this I mean either doing:

  - syslog-style (timestamp N identical messages suppressed)
  - client-side (with an expander next to a cluster of identical messages)

Note that syslog-style allows one to understand the length of time over which repeated messages arrived, which for a long process that is running is a non-trivial aid to debugging.

Folding of the very last log line is an interesting problem, as the entry displayed will actually change as repeated messages come up. Consider:

  timestamp A: message 1

Later:

  timestamp B: N identical messages supressed
  timestamp A: message 1

Later:

  timestamp C: N+M identical messages suppressed
  timestamp A: message 1

Even if there are performance issues being caused by large log entries, these are a) likely bugs, i.e. bug 1402237 and b) fixable through optional trucation of old entries, which seems sane to me.