Priority / Severity redux: vote!

[Adam Williamson]

Well, this got no response when posted as a reply to another thread, so
let's give it its own :)

To recap, we decided a while back it would be desirable to have triagers
set priority / severity when triaging new bugs, according to a
consistent system.

This was proposed to the development group, and received no "don't do
that, you crazy mad fool!"-type feedback. So we're good on that front.

Therefore, I intend to go ahead and add this to the standard triage
procedure soon - deadline end of this week, as it's all been agreed.

The remaining issues are:

* fully document the conventions for setting priority / severity
* agree on either the Beland / Williamson Method or the Cepl Method

on the second point - to recap, the Beland / Williamson Method is for
triagers to set both priority and severity, with severity indicating the
significance of the issue with regard only to the package it occurs in,
while priority indicates the significance of the issue with regard to
the distribution as a whole.

For e.g., a bug which caused all the icons in the Firefox toolbar to
disappear might be medium severity (it's not really a highly severe bug
*in Firefox*, there's no crashing or data loss) but urgent priority (as
it'd look really bad to release Fedora with such a visible bug in a
package almost everyone uses). A bug which caused some obscure
application that few people use to crash on startup would be high or
urgent severity (it's obviously a bad bug in the app), but low or medium
priority (it's not really that important to the project as a whole as
it's not widely used).

The Cepl Method would have triagers set only the severity. Low, medium
and high severity would indicate the importance of the issue just for
the package it was in, much like in the B / W Method. Urgent severity
would be used if it was a bad issue that was also, in the triager's
judgement, particularly significant for the distribution as a whole.
Priority would be left alone, on the understanding it should be used
solely by the maintainer.

In today's IRC meeting, most people came down in favour of the Cepl
Method, so unless there's a significant vote the other way here, we'll
go ahead and start implementing the Cepl Method at the end of this week.
If you have further input on this or would prefer the Beland /
Williamson method, speak up now!
-- 
Adam Williamson
Fedora QA Community Monkey
IRC: adamw | Fedora Talk: adamwill AT fedoraproject DOT org
http://www.happyassassin.net