On Wednesday, December 08, 2010 16:16:20 Andrei Alexandrescu wrote:
> On 12/8/10 3:25 PM, Jonathan M Davis wrote:
> > On Wednesday, December 08, 2010 15:03:43 Brad Roberts wrote:
> >> On Wed, 8 Dec 2010, Brad Roberts wrote:
> >>> I don't think we have concensus (or even a majority) that suggests moving forward.  I think it does do harm.  I think it suggests that we pay attention to votes, when reality suggests more strongly that we don't.
> >>> 
> >>> Fix the behavior first, imho.
> >> 
> >> Summary so far:
> >> 
> >> yes
> >> 
> >>   Andrei Alexandrescu
> >> 
> >> alternate suggestion
> >> 
> >>   David Simcha
> >> 
> >> no
> >> 
> >>   Brad Roberts
> >>   Jesse Phillips
> >>   Steve Schveighoffer
> >>   Don Clugston
> > 
> > I definitely like the idea of increasing the number of votes and making it possible to apply multiple votes to a bug. I definitely have more than 10 bugs that I'd like to vote on. But I do agree that there's little evidence that highly voted bugs get fixed. I don't think that I've ever had a bug that I've voted on fixed. And if the votes don't actually mean anything, then I'm not sure that there's much point in using them.
> > 
> > So, I like the idea of increasing the vote count, but if it doesn't affect what gets fixed, then perhaps it would be better to just get rid of the voting entirely. Based on votes, I would have expected stuff like the fact that Object isn't const-correct to have been fixed ages ago. But there are a number of bugs which have been around for quite a while which have a number of votes on them and yet never get fiixed.
> 
> The idea is to indeed have votes affect what is being worked on; otherwise I wouldn't bother you all. My problem is that I now must pawn votes from some bug to another. But the larger issue is that with few votes per user and many opened bugs we have essentially a bimodal distribution, whereas with more votes per user we have a longer, smoother tail, which is more informative.
> 
> I don't really understand the stiff opposition to this. This is not the time and the place to criticize the past and present of our process, but instead to take steps to improve it. Part of improving the process, now that fixing bugs is a major focus, is to figure out a good ranking function for bugs. And I believe that increasing the number of votes per user would contribute to that. Could we please push this through? Let's make it happen.

The change is fine with me. I just don't see much point if the votes are paid attention to, and the current impression seems to be that the votes aren't paid attention to. By the number of votes, it's fairly clear that not all that many people vote anyway. Perhaps changing the number of votes (particularly if you can then vote multiple times on bugs that matter more to you) will have a positive effect. But regardless, there does need to be at least some focus on bugs with a high vote count, or there's no point to having the votes.

I don't really see the change as hurting anything. I just think that if it's made that it needs to be followed up on by having the highly voted bugs actually given some kind of priority in terms of when they're fixed.

- Jonathan M Davis

On Wed, 8 Dec 2010, Andrei Alexandrescu wrote:

> On 12/8/10 4:30 PM, Brad Roberts wrote:
> > total open bugs: 2186
> > bugs with at least one vote: 328 (15%)
> > 
> > votes  num bugs
> >    0    1858
> >    1    120
> >    2    59
> >    3    17
> >    4    6
> >    5    7
> >    6    4
> >    7    3
> >    9    1
> >   10    1
> >   12    2
> >   13    1
> >   16    2
> >   17    1
> >   19    2
> >   22    1
> >   42    1
> > 
> > What sort of curve are you looking for?  What percentage of bugs with votes?
> 
> I'm hoping for a larger percentage of voted bugs and a smoother decay from the most voted bug to the least voted. Increasing the points available to each voter might improve both.

More data.  Looking at all votes placed (regardless of bug status):

| numvotes | numvoters |
+----------+-----------+
|        1 |        62 |
|        2 |        30 |
|        3 |        14 |
|        4 |        10 |
|        5 |         9 |
|        6 |         4 |
|        7 |         8 |
|        8 |         5 |
|        9 |        10 |
|       10 |        40 |

total numvoters: 192 (of 863 users, so 22%)

I find this data interesting.  An awful lot of people (79% of the people who vote, or 95% of the total reporting population) aren't using all the votes they're currently given.

There are 3145 bugs that are marked resolved or closed, 90 of which have had at least one vote.  Break down of number of votes per bug for the resolved bugs.

| numvotes | numbugs |
+----------+---------+
|       11 |       1 |
|        7 |       1 |
|        6 |       2 |
|        5 |       6 |
|        3 |       3 |
|        2 |      13 |
|        1 |      64 |

Anyone else have any other statistical data they'd like pulled from the db?

I've gone ahead and increased the number of allowed votes to 100 with up to 10 votes for each issue.  I'm tired of debating it and it's really not that important to me.

Later,
Brad

On Thu, Dec 9, 2010 at 10:44 AM, Brad Roberts <braddr at puremagic.com> wrote:

> On Wed, 8 Dec 2010, Andrei Alexandrescu wrote:
>
>
>
> I find this data interesting.  An awful lot of people (79% of the people who vote, or 95% of the total reporting population) aren't using all the votes they're currently given.
>

Personally, with only 10 votes I'm highly conservative in voting. I only
vote for the bugs that I think are critical. With more votes,
I'll use more to vote for other bugs, such as those I find frustrating or
annoying, but aren't showstoppers. Or those that I haven't
yet encountered and don't ever want to. If others think like me, and Walter
& Andrei & Co. use the new rush of votes to prioritize
bugfixes, then this can only be a good thing.


-- 
Mike Parker
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.puremagic.com/pipermail/phobos/attachments/20101209/71e96fbf/attachment-0001.html>

...which is why I recommend allowing multiple votes per bug.

On Wed, Dec 8, 2010 at 10:24 PM, Michael Parker <aldacron at gmail.com> wrote:

> On Thu, Dec 9, 2010 at 10:44 AM, Brad Roberts <braddr at puremagic.com>wrote:
>
>> On Wed, 8 Dec 2010, Andrei Alexandrescu wrote:
>>
>>
>>
>> I find this data interesting.  An awful lot of people (79% of the people who vote, or 95% of the total reporting population) aren't using all the votes they're currently given.
>>
>
> Personally, with only 10 votes I'm highly conservative in voting. I only
> vote for the bugs that I think are critical. With more votes,
> I'll use more to vote for other bugs, such as those I find frustrating or
> annoying, but aren't showstoppers. Or those that I haven't
> yet encountered and don't ever want to. If others think like me, and Walter
> & Andrei & Co. use the new rush of votes to prioritize
> bugfixes, then this can only be a good thing.
>
>
> --
> Mike Parker
>
> _______________________________________________
> phobos mailing list
> phobos at puremagic.com
> http://lists.puremagic.com/mailman/listinfo/phobos
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.puremagic.com/pipermail/phobos/attachments/20101208/d70229e4/attachment.html>

On Thu, 9 Dec 2010, Michael Parker wrote:

> On Thu, Dec 9, 2010 at 10:44 AM, Brad Roberts <braddr at puremagic.com> wrote:
> 
> > On Wed, 8 Dec 2010, Andrei Alexandrescu wrote:
> >
> >
> >
> > I find this data interesting.  An awful lot of people (79% of the people who vote, or 95% of the total reporting population) aren't using all the votes they're currently given.
> >
> 
> Personally, with only 10 votes I'm highly conservative in voting. I only
> vote for the bugs that I think are critical. With more votes,
> I'll use more to vote for other bugs, such as those I find frustrating or
> annoying, but aren't showstoppers. Or those that I haven't
> yet encountered and don't ever want to. If others think like me, and Walter
> & Andrei & Co. use the new rush of votes to prioritize
> bugfixes, then this can only be a good thing.
> 
> 
> -- 
> Mike Parker
> 

Votes aren't immutable.  Assign 'em all and then move them as something higher priority rolls around.
-------------- next part --------------
_______________________________________________
phobos mailing list
phobos at puremagic.com
http://lists.puremagic.com/mailman/listinfo/phobos

On 9 December 2010 01:16, Andrei Alexandrescu <andrei at erdani.com> wrote:
> I don't really understand the stiff opposition to this. This is not the time and the place to criticize the past and present of our process, but instead to take steps to improve it. Part of improving the process, now that fixing bugs is a major focus, is to figure out a good ranking function for bugs. And I believe that increasing the number of votes per user would contribute to that. Could we please push this through? Let's make it happen.


Let me explain the stiff opposition.

This change DESTROYS THE EXISTING STATISTICS.
When you make a radical change like this, you have to wait for at a
least a year or so before the statistics are meaningful again.
It took a couple of years to get them to the level they are at now. Seriously.
As one of the very few people who actually *uses* the vote statistics,
to determine which bugs to fix,  I'm quite annoyed about this.


You also made this statement on the newsgroup:
> Right now the process of choosing which bug to fix next is unstructured.

I think this indicates a misconception.
Over the last six months, nearly 90% of fixed bugs were patches. I
think it's entirely appropriate that bugs with correct patches get
high priority.
And since they are community-driven, there's not  much control over
which bugs they are.

Half the patches are mine. I follow a definite priority in choosing
which bugs to patch.
1.  Wrong-code regression
2.  ICE regression
3.  wrong code bugs marked 'blocker' or 'critical'
4. Compiler segfault
5. Missing line number
6. ICE or wrong code, non-obscure situation
7. Regression, non-obscure (most recently introduced get highest priority)
8. ICE or wrong code, more obscure situation

Bugs with votes jump a couple of places up the list. There *ought* to be categories below those I've listed, where votes and age would play a big role,  but there have just been too many bugs in the top priority categories.

I also patch a few bugs which aren't on that list, mostly when they're very easy, or when I hit them myself. I also do CTFE bugs, because I feel I've taken ownership of that part of the compiler.

Note that people don't tend to vote for ICE and wrong-code bugs, because you don't notice them until after they've wasted days of your life.

Hm... I meant the Normal/Critical/Blocker thingy.  If that's not priority, then I meant whatever that field is.

But I would consider bugs marked as blockers more severe and more urgent.

I don't think the voting system will reflect the exact pain levels of the person.  People exaggerate.  For example, those surveys where you have 1-5, 1 = not likely at all, and 5 = definitely, people often circle all 5s or all 1s because they want to drag the curve in a certain direction.  I'd likely only vote 10 votes on a bug or none at all, because I don't know how to really measure my pain per bug.  Others who only want one bug fixed will vote 100 for that bug not because it's really painful but because they don't care about any other bugs.

Giving people one vote per bug seems reasonable.  It's basically saying, out of the hundreds or thousands of bugs, these are my top 10.  Like Brad says, I think we should ensure we use the voting system to guide the direction of bug fixing before we try a new voting system.  Also, if you keep voting limited like it is now, bugs with a large number of votes are likely to be major points of pain because they affect many people.

Can we have a published list somewhere of the most voted bugs?  There used to be the monthly "most requested features" of D, which was horrifically out of date, but it reminded us every month that it was there.  A monthly post to the NG of the top 10 or top 20 most voted bugs would help keep people focused on the votes IMO.

-Steve



----- Original Message ----
> From: Andrei Alexandrescu <andrei at erdani.com>
> To: Discuss the phobos library for D <phobos at puremagic.com>
> Sent: Wed, December 8, 2010 5:15:24 PM
> Subject: Re: [phobos] Upping the number of bugzilla votes from 10 to 100
> 
> The P1-P5 thing? Never knew which is most severe :o).
> 
> Andrei
> 
> On  12/8/10 1:10 PM, Steve Schveighoffer wrote:
> > Bugs have a severity field.  I'd say use that.
> >
> > -Steve
> >
> >
> >      *From:* David Simcha <dsimcha at gmail.com>
> >      *To:* Discuss the phobos library for D <phobos at puremagic.com>
> >      *Sent:* Wed, December 8, 2010 4:04:34 PM
> >      *Subject:* Re: [phobos] Upping the number of bugzilla votes from  10
> >     to 100
> >
> >     If we want a more  analog system, I'd say give every user 100 votes
> >     and allow  more than one to be used on a single bug. This way, bugs
> >      that are a huge PITA can can be given a lot of votes and ones that
> >      are a little annoying can be given a few, instead of each  user
> >     having to make an all-or-nothing  decision.
> >
> >
> >
> >
> >  _______________________________________________
> > phobos mailing  list
> > phobos at puremagic.com
> >  http://lists.puremagic.com/mailman/listinfo/phobos
> _______________________________________________
> phobos  mailing list
> phobos at puremagic.com
> http://lists.puremagic.com/mailman/listinfo/phobos
> 

On Thu, Dec 9, 2010 at 7:47 AM, Steve Schveighoffer <schveiguy at yahoo.com>wrote:

> Hm... I meant the Normal/Critical/Blocker thingy.  If that's not priority,
> then
> I meant whatever that field is.
>
> But I would consider bugs marked as blockers more severe and more urgent.
>
> I don't think the voting system will reflect the exact pain levels of the
> person.  People exaggerate.  For example, those surveys where you have 1-5,
> 1 =
> not likely at all, and 5 = definitely, people often circle all 5s or all 1s
> because they want to drag the curve in a certain direction.  I'd likely
> only
> vote 10 votes on a bug or none at all, because I don't know how to really
> measure my pain per bug.  Others who only want one bug fixed will vote 100
> for
> that bug not because it's really painful but because they don't care about
> any
> other bugs.
>
> Giving people one vote per bug seems reasonable.  It's basically saying,
> out of
> the hundreds or thousands of bugs, these are my top 10.  Like Brad says, I
> think
> we should ensure we use the voting system to guide the direction of bug
> fixing
> before we try a new voting system.  Also, if you keep voting limited like
> it is
> now, bugs with a large number of votes are likely to be major points of
> pain
> because they affect many people.
>
> Can we have a published list somewhere of the most voted bugs?  There used
> to be
> the monthly "most requested features" of D, which was horrifically out of
> date,
> but it reminded us every month that it was there.  A monthly post to the NG
> of
> the top 10 or top 20 most voted bugs would help keep people focused on the
> votes
> IMO.
>

You mean post the results of this query every month: http://goo.gl/xDqOe? It just sounds like adding clutter to the newsgroup to me.

I don't think it's not hard to query the issue tracker if someone wants to work on fixing a bug.

jcc7
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.puremagic.com/pipermail/phobos/attachments/20101209/b744e7b6/attachment.html>

>
>From: Justin C Calvarese <technocrat7+d at gmail.com>
>To: Discuss the phobos library for D <phobos at puremagic.com>
>Sent: Thu, December 9, 2010 9:43:23 AM
>Subject: Re: [phobos] Upping the number of bugzilla votes from 10 to 100
>
>On Thu, Dec 9, 2010 at 7:47 AM, Steve Schveighoffer <schveiguy at yahoo.com>
wrote:
>
>
>>Can we have a published list somewhere of the most voted bugs?  There used to
>be
>>the monthly "most requested features" of D, which was horrifically out of
date,
>>but it reminded us every month that it was there.  A monthly post to the NG of the top 10 or top 20 most voted bugs would help keep people focused on the
>votes
>>IMO.
>>

You mean post the results of this query every month: http://goo.gl/xDqOe? It just sounds like adding clutter to the newsgroup to me.

I don't think it's not hard to query the issue tracker if someone wants to work on fixing a bug.

Well, I think giving the community feedback as to what bugs are listed as higher priority without requiring them to manipulate the bugzilla interface would help people to remember to vote, plus it gives the developers a reminder as to what needs fixing.

In my company, we have a weekly meeting discussing all open issues, and it was a way better tool to get people to fix issues than just hoping they looked at the bug list periodically.  In effect, posting to the NG is like a squeaky wheel calling for grease.

And is the NG really so uncluttered that it can't have one more post per month? ;)

-Steve




-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.puremagic.com/pipermail/phobos/attachments/20101209/97db2505/attachment-0001.html>

On 9 December 2010 14:47, Steve Schveighoffer <schveiguy at yahoo.com> wrote:
> Giving people one vote per bug seems reasonable. ?It's basically saying, out of the hundreds or thousands of bugs, these are my top 10. ?Like Brad says, I think we should ensure we use the voting system to guide the direction of bug fixing before we try a new voting system. ?Also, if you keep voting limited like it is now, bugs with a large number of votes are likely to be major points of pain because they affect many people.

I agree. But I think that increasing the limit to (say) 20 bugs per
person would be OK.
Not many people would use it, and the existing statistics would remain valid.

> Can we have a published list somewhere of the most voted bugs? ?There used to be the monthly "most requested features" of D, which was horrifically out of date, but it reminded us every month that it was there. ?A monthly post to the NG of the top 10 or top 20 most voted bugs would help keep people focused on the votes IMO.

Who do you mean by "people"? Do you mean "Walter"?

> In effect, posting to the NG is like a squeaky wheel calling for grease.

Yes. But where the hell is the grease? It's not actually created by the squeaking, unfortunately.

The problem is that there are 70 bugs/month fixed, but >150 bugs reported/month. (incidentally 35/month come from bearophile).