672191 - Guided bug form should set OS to match the user's agent string. (Currently submits with OS=other)

Status: RESOLVED FIXED

(bugzilla.mozilla.org :: Extensions, defect)

Description

[Wayne Mery (:wsmwk)]

Guided bug form should set OS to match the user's agent string.  

I don't know if it happens in all cases, i.e. what circumstances, because I did not test, but many of our bugs are being filed as OS=other -  http://tinyurl.com/43wo37w  which is confusing, and affects searchability

Comment 1

[:glob ✱]

the guided form doesn't make any attempt to guess the o/s on purpose; they're far more often wrong than right, and they don't apply to most of our products.  it's a bug that it selects "Other" instead of "All", and i'll address that immediately.



one possibility is to set the platform and OS automatically, but only for a small subset of products, such as firefox, thunderbird, fennec and perhaps core.

Comment 2

[:glob ✱]

/me fixes the wrong platform and os on this bug, because it was automatically detected.

Comment 3

[Wayne Mery (:wsmwk)]

(In reply to comment #1)
> It's a bug that it selects "Other" instead of "All", and i'll
> address that immediately.

Thanks. I'm changing the summary here to emphasize the desire for OS = agent string, so please fix "Other" on a separate bug. 


> they're far more often wrong than right, and they don't apply to most of our
> products.
> ...
> one possibility is to set the platform and OS automatically, but only for a
> small subset of products, such as firefox, thunderbird, fennec and perhaps
> core.

"right or wrong" can be a matter degrees, and perspective. (and so I hate this bug report for that reason)

From the perspective of this triager, it is more useful for the initial report to be OS = the OS of the reporter and be "wrong" at least initially in terms of "is the bug really applicable to all OS?" (pending further triage) ... than to have it set to ALL and not be able to efficiently find potentially OS specific bugs.  

For example, if I want to find reports for a type of bug that I "suspect" is only happening on Mac, X months ago I could create a query to do that.  Yes, I can (and do) accept that the results will not be 100% accurate, but they used to be pretty darn good.  Today, the same query is less accurate and will increasingly be so.  

Realistically, after the initial bug report, how many devs and triagers are going to change an OS=ALL bug to OS=<something else>?  There will be some - bug Bug 672189 is an example. But it will be rare.

Those who argue the contrary (rationally of course), and want OS=ALL, often assume or cite such things as :
a) no one queries on OS (false)
b) all or most bugs get triaged and therefore OS is accurate (but almost no component/product achieves any high level of accuracy)
c) reporters and devs don't care about the OS, therefore it is frequently not accurately set (true), plus a waste of a person's time (a matter of perspective)
d) we have two camps (devs and users alike) of interpretation of what OS should mean, 1. OS should be the REPORTER'S OS and should never change, and 2. OS should equal the universe of users affected (no one seems to agree who is "right")
e) dev (those who fix bugs) perspective seems to most often be that it doesn't matter what OS is set (and from from the perspective of FIXING, not triaging bugs, this is quite reasonable)

And no doubt there are more viewpoints.

But my viewpoint is primarily from that of being able to query, and OS=ALL makes doing (some) queries harder.

Comment 4

[:glob ✱]

(In reply to comment #3)
> Thanks. I'm changing the summary here to emphasize the desire for OS = agent
> string, so please fix "Other" on a separate bug. 

no worries, it wasn't a hijack attempt :)
 
> "right or wrong" can be a matter degrees, and perspective. (and so I hate
> this bug report for that reason)
> [snip]

i'm glad to fine tune things based on feedback, and you've been most helpful.

> d) we have two camps (devs and users alike) of interpretation of what OS
> should mean, 1. OS should be the REPORTER'S OS and should never change, and
> 2. OS should equal the universe of users affected (no one seems to agree who
> is "right")

to me this is the crux of the issue.  bug 579089 comment 13 has a proposal which will make this distinction clear, and to assist triage we could extend the current bug search, or develop a custom report which uses the user-agent information in comment 0 to track the reporter's os and platform.  (see https://bugzilla.mozilla.org/page.cgi?id=triage_reports.html for an example of a custom report).


until anything of that nature happens it sounds reasonable to restore os detection to firefox, thunderbird, and fennec bugs only, and i'll work on that.  i'm interested in your thoughts on if core bugs should be added to that list.

Comment 5

[:glob ✱]

i've backported the fix from bug 307688 and i've used that to set the default platform and op_sys for the following products: firefox, fennec, seamonkey, camino, core, and thunderbird.