Date:   Tue, 10 Mar 2020 09:40:43 -0400
From:   Josef Bacik <josef@...icpanda.com>
To:     Michal Hocko <mhocko@...nel.org>
Cc:     lsf-pc <lsf-pc@...ts.linuxfoundation.org>,
        Linux FS Devel <linux-fsdevel@...r.kernel.org>,
        linux-mm@...ck.org, linux-xfs@...r.kernel.org,
        Btrfs BTRFS <linux-btrfs@...r.kernel.org>, bpf@...r.kernel.org,
        linux-ext4@...r.kernel.org, linux-block@...r.kernel.org
Subject: Re: [LSFMMBPF TOPIC] Killing LSFMMBPF

On 3/10/20 9:13 AM, Michal Hocko wrote:
> On Fri 06-03-20 09:35:41, Josef Bacik wrote:
>> Hello,
>>
>> This has been a topic that I've been thinking about a lot recently, mostly
>> because of the giant amount of work that has been organizing LSFMMBPF.
> 
> There is undoubtedly a lot of work to make a great conference. I have hard
> time imagine this could be ever done without a lot of time and effort on
> the organizing side. I do not believe we can simply outsource a highly
> technical conference to somebody outside of the community. LF is doing a
> lot of great work to help with the venue and related stuff but content
> wise it is still on the community IMHO.
> 
> [...]
>> These are all really good goals, and why we love the idea of LSFMMBPF.  But
>> having attended these things every year for the last 13 years, it has become
>> less and less of these things, at least from my perspective.  A few problems
>> (as I see them) are
>>
>> 1) The invitation process.  We've tried many different things, and I think
>> we generally do a good job here, but the fact is if I don't know somebody
>> I'm not going to give them a very high rating, making it difficult to
>> actually bring in new people.
> 
> My experience from the MM track involvement last few years is slightly
> different. We have always had a higher demand than seats available
> for the track. We have tried really hard to bring people who could
> contribute the most requested topic into the room. We have also tried to
> bring new contributors in. There are always compromises to be made but
> my recollection is that discussions were usually very useful and moved
> topics forward. The room size played an important role in that regard.
> 
>> 2) There are so many of us.  Especially with the addition of the BPF crowd
>> we are now larger than ever.  This makes problem #1 even more apparent, even
>> if I weighted some of the new people higher who's slot should they take
>> instead?  I have 0 problems finding 20 people in the FS community who should
>> absolutely be in the room.  But now I'm trying to squeeze in 1-5 extra
>> people.  Propagate that across all the tracks and now we're at an extra
>> 20ish people.
> 
> Yes, BPF track made the conference larger indeed. This might be problem
> for funding but it didn't really cause much more work for tracks
> organization (well for MM at least).
> 
>> 3) Half the people I want to talk to aren't even in the room.  This may be a
>> uniquely file system track problem, but most of my work is in btrfs, and I
>> want to talk to my fellow btrfs developers.  But again, we're trying to
>> invite an entire community, so many of them simply don't request
>> invitations, or just don't get invited.
> 
> I do not have the same experience on the MM track. Even though the whole
> community is hard to fit into the room, there tends to be a sufficient
> mass to move a topic forward usually. Even if we cannot conclude many
> topics there are usually many action items as an outcome.
> 
> [...]
> 
>> So what do I propose?  I propose we kill LSFMMBPF.
> 
> This would be really unfortunate. LSFMMBPF has been the most attractive
> conference for me exactly because of the size and cost/benefit. I do
> realize we are growing and that should be somehow reflected in the
> future. I do not have good answers how to do that yet unfortunately.
> Maybe we really need to split the core agenda and topics which could be
> discussed/presented on other conferences. Or collocate with another
> conference but I have a feeling that we could cover more since LSFMMBPF
> 

LSFMMBPF is still by far the most useful conference I attend, so much so that 
it's basically the only thing I attend anymore.

My point is less about no longer having a conference at all, and more about 
changing what we currently have to be more useful to more people.  For MM, and I 
assume BPF, it's much different as you guys are all on the same codebase.  You 
get 25 people in the room chances are a much larger percentage of you are 
interested in each individual topic.

File systems and storage?  Way less so.  We've expanded to 3 days of conference, 
which has only exacerbated this issue for me.  Now I have a full day that I'm 
trying to fill with interesting topics that we're all interested in, and it's a 
struggle.  If instead we had everybody from the file system community there then 
I could just say "OK day 3 is BoF day, have your FS specific meetups!" and be 
done with it.  But as it stands I know XFS is missing probably 1/3 of their main 
contributors, and Btrfs is missing 1/2 to 2/3 of our developers.

In order to accomplish that we need to radically change the structure of the 
conference, hence my hyperbolic suggestion.  I think what Ted suggested is 
probably my ideal solution, we have a kernel focused spring conference where the 
whole community gets together, and then we have tracks that we carve up.

But is it a problem worth solving?  I'm not sure.  I know how I feel, but maybe 
I'm the crazy one.  I think its worth discussing.  If more people like how we 
currently do it then we can just keep trucking along.  It's not like I'll stop 
showing up, this is still a tremendously useful conference.  I just think we can 
do better.  Thanks,

Josef

Powered by blists - more mailing lists