Last rites
Posted by wolfger on May 7, 2008
When I gave my presentation on Ubuntu Bug Triage Basics at Penguicon (or whenever anybody asks what triage is), I likened bug triage to the TV show M*A*S*H in which the wounded come in, go through triage, and then are operated on according to their triaged status. While the analogy between software bug triage and human triage isn’t perfect, it’s fairly close.
Lately, I’ve been feeling a lot like Father Mulcahy. As the Army minister of the 4077th, he had the unenviable task of performing last rites on those who could not be saved. In much the same fashion, I’ve been invalidating bug reports that have sat around for several months in need of more information. If we don’t have enough information, we can’t operate. If we are unable to operate for too long, the bug report dies. It may not seem like the most useful triage function, but I like to do it for two reasons:
- It’s quick and (usually) simple. I can crunch out a lot of these is a small amount of time.
- It’s a good way to decrease the number of open bugs, of which there are way too many for our team of dedicated (mostly volunteer) developers to address them all.
In addition to those obvious reasons, though, there’s some other good that comes out of these “last rites”. Sometimes (in fact, twice in the past two days) the bug filer will get my invalidation notice, and suddenly wake up, see the questions that have been needing answers, and answer them! Kind of like that one episode of M*A*S*H… “This one’s not dead!”
And this, to answer a question from the Penguicon presentation, is another reason we don’t just script the process of invalidating old incomplete bugs.
Sarah Elkins said
“Sometimes (in fact, twice in the past two days) the bug filer will get my invalidation notice, and suddenly wake up, see the questions that have been needing answers, and answer them! Kind of like that one episode of M*A*S*H… “This one’s not dead!””
Hurray! A happy ending.
If the bug invalidation process were scripted, couldn’t it still notify the original filer and give an email address (person or list) for further discussion?
wolfger said
It could, and it would probably work almost as well in that case, but scripted replies tend to look and feel scripted, and are less likely to gain the human attention that a more personalized note can get.
The recent hubbub in the bug squad about devs not wanting us to invalidate their stuff (even when it’s left “incomplete” for a half year) is another case against scripts…. Sure, you can script around that too, but the complexity increases, and as we find more and more edge cases, one begins to wonder if the script is actually going to do the job right. If you’ve got to double-check the script’s work, why have a script?