launchpad-dev team mailing list archive

[Robert Collins <robertc@xxxxxxxxxxxxxxxxx>]

On Sat, Feb 18, 2012 at 7:55 AM, Deryck Hodge
<deryck.hodge@xxxxxxxxxxxxx> wrote:

> It's not released, though.  Only lp devs see it.  Most of you heavy
> tab-using zealots have been sufficiently afflicted and know better
> now. :)  Personally, I think it's good for you.  Multi-tasking is a
> lie, you 300 tab loving dev you. ;)
>
> But seriously, it's a problem that has limited impact (in terms of who
> it is affecting), and once you know, you can work around it.

Folk who know are repeatedly forgetting and consuming scarce resources
doing that. This is little different to any other foot gun we might
have, which we'd rollback until we're ready to do it more smoothly.

> I view it as a work in progress that isn't complete.  Not as stalled work.
>
> Even if you disagree, I think there's a whole spectrum between "we
> *must* resource it *now*" or "we *must* disable it *now*".  I guess
> I'm arguing for a discussion about how to resource and that it's
> better enabled than disabled until we can resource it. But I take your
> points that it's incomplete and not ready for release. I agree on
> that.

Francis and I talked on IRC, he believes it is maintenance turf (and I
agree - there is little reason to believe it would grow in scope); the
bug list I linked before was wrong -
https://bugs.launchpad.net/launchpad-project/+bugs?field.tag=longpoll
is the right one.

There are two major reasons to disable this (irrespective of it being
worked on):
 - it is at best breakeven for us today, not yet a benefit. reduces
dev friction, increases ops friction. Increases dev confusion.
 - we are seeing a different Launchpad to our users - and not a
difference like our debug tools, but a difference in terms of features
and usability.

If the thing was being worked on right now, we could say 'lets
accommodate these reasons for a week so that the developers working on
this thing can dogfood' or even 'this is early alpha testing we need
that feedback'.

But neither is the case: we've learnt what we need to know (that
concurrent connections from tabs are an issue), we know what remains
to move the core infrastructure to a supported production ready state
: we're not learning anything by having the feature enabled. And we're
paying.

-Rob