Re: No Issue Tracker - Say it Ain't So! - Mailing list pgsql-hackers

On Wed, Sep 30, 2015 at 12:43 PM, Josh Berkus <josh@agliodbs.com> wrote:
> On 09/30/2015 07:44 AM, Merlin Moncure wrote:
>> I'm not trolling in any way.  I'm just challenging you to back up your
>> blanket assertions with evidence.  For example, you're assertion that
>> mailing lists are insufficient is simply stated and expected to be
>> taken on faith: *How* is it insufficient and *what* do things like in
>> the new world?  Be specific: glossing over these details doesn't
>> really accomplish anything and avoids the careful examination that may
>> suggest small tweaks to the current processes that could get similar
>> results with a lot less effort.  In this entire massive thread, so far
>> only Josh has come up with what I'd consider to be actionable problem
>> cases.
>
> I don't see any way to make small tweaks to the existing process which
> would fix any of these problems.  I think if that were possible, we'd
> already have done it.  Suggestions welcome, of course.
>
> For example, "just use the wiki for this" has been mentioned as an
> alternative.  But we've tried "just using the wiki" for a number of
> things, and it doesn't really work.  For example, using the wiki as a
> way of breaking down the various multixact issues manifestly didn't
> work.  A big part of the problem there is that there's no good way for
> the wiki to notify people when there's been an update; a smaller part is
> that the formatting gets messed up and impossible to follow.

Agree that wiki are mainly suited for documentation.  But they can
also be used to prototype new processes in a hurry or at least sketch
them out.  This is BTW how the commit fest application spawned up more
or less (which I happen to think works quite well).

But this brings up a couple of points and more questions:
*) Every wiki software I've seen, including mediawiki, allows
subscription to pages for changes.  This is pretty much the same model
most issue trackers use BTW except maybe the automatic subscription
rules are a little different.

*) Given that, I'm not sure that issue trackers are really an
improvement on that point.  In fact, fragmentation of communication is
a central complaint I have with them, including JIRA since people have
to subscribe and/or search for things of interest.  Of course, you can
configure them to just always send an email upon every response, but
then I'm thinking, 'why not just use email?'.  Also, I find the
suggestion that any $tool has a better search facility better than
google's to be laughable, except for very structured searches like
'issue type X by version Y'.

*) I only followed the multi-xact saga very loosely, except to see a
lot of angst for a significant bug (or, set of bugs) slipping out.
Are we sure that this problem didn't in fact stem from insufficient
review and/or testing?  And if not, how did using the wiki and/or not
having individual subtasks run through a tracker contribute to the
problem or its handling?  In other words, what does "manifestly didn't
work" mean?

I guess I'm yammering on here, but between us I'd suck the honey out
of an active beehive to have a day job dev process that more closely
resembles what this project uses, and I frequently exclaim so to
whomever is unfortunate enough to walk by.  But the tool brandishing
zealots have taken over, and I spend a non-trivial of the day filling
out various forms and re-explaining things over and over  (and the
rules are even then much relaxed for me, because I'm, you know,
"special").

My take is that this project is pretty well run and has taught me the
credo, "people and processes first, tools second".  The gripes raised
so far seem awfully vague for the most part, TBH.

merlin