[Mesa-dev] XDC 2017 feedback
Daniel Stone
daniel at fooishbar.org
Thu Sep 28 08:11:06 UTC 2017
Hi,
On 28 September 2017 at 03:49, Ian Romanick <idr at freedesktop.org> wrote:
> On 09/27/2017 04:55 PM, Rob Clark wrote:
>> Sadly by the time we were aware of the dates for the khronos f2f it
>> was not possible to change the dates for XDC :-(
>>
>> The XDC dates were set in Feb, and afaict the khronos dates were
>> announced in July (?), so take this up with khronos ;-)
>
> Ok... so we're going to go there.
>
> Frankly, that's a giant steaming load of bull.
That's a bit much; can you please tone it down?
> Blocks of hotel rooms,
> multiple conference rooms for 500+ people, and catering for the Khronos
> meeting was booked in late 2016. We're already working on contracts for
> the September 2018 meeting. Contracts of this scale are really hard to
> change. There are 5x to 10x as many people at a Khronos face-to-face as
> at XDC. Events of that scale have a massively deep pipeline.
>
> Google was just unwilling to find a different dates for space *at their
> own campus*. That's really, really weak. This is especially
> infuriating because there are numerous Googlers who attend the Khronos
> meetings. Did the organizers poll any of them? The XDC organizers
> clearly did not even exercise due diligence to detect a possible
> conflict. If the organizers had cared to be aware of dates of
> conflicting events, they would have known.
For starters, yes, it obviously sucks that the two clashed. That being said ...
Logistically, Khronos (and their multiple full-time paid
administrative staff) can move on two axes: time and location.
Location is easy, since any city of any size has a number of four-star
hotels who can host that number of people in exchange for eyewatering
amounts of money directly, as well as guaranteed blocks of rooms at
rates many XDC attendees couldn't afford. With the hosting locked down
to avoid the insane cost, we could only move XDC in time rather than
space.
Time-wise, it's not just Khronos F2F which, being quarterly, is pretty
hard to avoid no matter when you do it. There was Plumbers and OSS NA
before (with the no-go zone of school holidays before that, and
European holidays in August), and afterwards Linaro Connect, Kernel
Recipes, ELCE, GNOME/Qt conferences, the GStreamer conference, and
Kernel Summit. That's before you even touch things like IBC. It really
hurts time-wise, and choosing not to clash means going directly
back-to-back (asking people to be away for 2 weeks at a time), or you
put space between them and ask people to do intercontinental trips
twice in three weeks. At least XDC managed to not overlap the Vulkan
WSI sessions at the F2F: there were people at XDC who'd done both.
I don't know why you think an event of 105 people doesn't have a deep
pipeline either ... ?! Most places don't have a surfeit of 120-person
meeting rooms. The ones they do have tend to book up a long time in
advance, for obvious reasons. Given that we were mostly external,
those rooms have to be not in sensitive buildings, and on the ground
floor close to an exit. Even with that, it required two full-time
dedicated security staff (another logistical dependency) to make sure
the herds of external people didn't end up wandering through an
otherwise badge-only building. The catering also needed to be booked,
and A/V staff to assist. We had a dependency on Jen and Stephane's
time as the local support: they had to be there in person, and be able
to dedicate their full week to the conference.
None of that is being 'unwilling', 'weak', 'not exercis[ing] due
diligence', or not caring. It's organising a conference actually being
really difficult: a deference you're paying Khronos (who have an
objectively easier job), and not the people who organise XDC. Just it
doesn't work out perfectly, doesn't mean it's due to idiocy or lack of
effort.
Even if you do get all your ducks perfectly in a row, sometimes
Network Rail announce a few weeks after you've booked everything that
they're going to close the line to Cambridge for engineering works, so
it's suddenly really difficult to even get there in the first place.
Ho hum.
Hopefully next year it doesn't clash. But I can guarantee you that
even if it doesn't, there will still be people who are unable to come,
because there is just no globally-optimal solution.
Cheers,
Daniel
More information about the mesa-dev
mailing list