Hello everyone,


Following a number of discussions with Walter regarding lifetime management for class objects, we've had a small epiphany: we need to frame collaboration better. Let me explain.

There are a number of very good folks here with type system-y stuff: Timon, Amaury, Michel, and more. For all I know each of them is better than I could be. In the past Walter and I would show DIP74 and people would point out mistakes and insufficiencies in it. That was suboptimal; it has put potential collaborators in a contrarian, negative position. Instead, we should co-opt our talented contributors to do positive work, not spend time countering inferior work.

For what I know Amaury is working alone on a DIP regarding ownership types. That's great, but I think we could do even better by helping a more collaborative environment.

So by this I'm asking folks' participation to a study group on object lifetime. Walter and I believe we must make reference counting for classes work, and that Rust-style schemes are too complex for their own good. But we're very willing to be convinced otherwise.

Normally we'd be holding this on the forum, but as we all know, forum discussions tend to meander a lot and lose focus. For that reason, please write me email about joining a mailing list dedicated to discussions on object lifetime - a study group. The outcome should be a DIP that moves forward safe lifetime management for D.


Thanks,

Andrei

On 27/10/15 3:45 PM, Andrei Alexandrescu wrote:
> Hello everyone,
>
>
> Following a number of discussions with Walter regarding lifetime
> management for class objects, we've had a small epiphany: we need to
> frame collaboration better. Let me explain.
>
> There are a number of very good folks here with type system-y stuff:
> Timon, Amaury, Michel, and more. For all I know each of them is better
> than I could be. In the past Walter and I would show DIP74 and people
> would point out mistakes and insufficiencies in it. That was suboptimal;
> it has put potential collaborators in a contrarian, negative position.
> Instead, we should co-opt our talented contributors to do positive work,
> not spend time countering inferior work.
>
> For what I know Amaury is working alone on a DIP regarding ownership
> types. That's great, but I think we could do even better by helping a
> more collaborative environment.
>
> So by this I'm asking folks' participation to a study group on object
> lifetime. Walter and I believe we must make reference counting for
> classes work, and that Rust-style schemes are too complex for their own
> good. But we're very willing to be convinced otherwise.
>
> Normally we'd be holding this on the forum, but as we all know, forum
> discussions tend to meander a lot and lose focus. For that reason,
> please write me email about joining a mailing list dedicated to
> discussions on object lifetime - a study group. The outcome should be a
> DIP that moves forward safe lifetime management for D.
>
>
> Thanks,
>
> Andrei

Is it possible to have a read only interface/receiving?
Because I'm interested in the content, but not enough knowledge to actually talk about it.

On Tuesday, 27 October 2015 at 02:45:24 UTC, Andrei Alexandrescu wrote:
> So by this I'm asking folks' participation to a study group on object lifetime. Walter and I believe we must make reference counting for classes work, and that Rust-style schemes are too complex for their own good. But we're very willing to be convinced otherwise.
>

+1, but don't forget about structs.
With inc/dec elision, most of the overhead of RC is avoided anyways - I'll see if I can find the paper where someone did it for a common lisp implementation and it reduced inc/dec counts by 90% or some-such.

On Tuesday, 27 October 2015 at 02:56:56 UTC, Rikki Cattermole wrote:
> Is it possible to have a read only interface/receiving?
> Because I'm interested in the content, but not enough knowledge to actually talk about it.

+1

On Tuesday, 27 October 2015 at 02:45:24 UTC, Andrei Alexandrescu wrote:
>
> Normally we'd be holding this on the forum, but as we all know, forum discussions tend to meander a lot and lose focus. For that reason, please write me email about joining a mailing list dedicated to discussions on object lifetime - a study group. The outcome should be a DIP that moves forward safe lifetime management for D.

+1, but creating a dedicated mailing list feels exclusive to me again. I would at least like to read what they are talking about once a while. Maybe mark the threads "experts only" or something.

Invite-only mail list with archive being published via forum.dlang.org interface (read-only) sounds like a best match to be.

I agree that public NG is not very suitable for any serious technical case study, this has been the recurring pattern.

On 10/26/15 10:56 PM, Rikki Cattermole wrote:
> Is it possible to have a read only interface/receiving?
> Because I'm interested in the content, but not enough knowledge to
> actually talk about it.

Then you may want to join and opt not to post. -- Andrei

On 10/27/15 4:42 AM, Robert burner Schadek wrote:
> On Tuesday, 27 October 2015 at 02:45:24 UTC, Andrei Alexandrescu wrote:
>>
>> Normally we'd be holding this on the forum, but as we all know, forum
>> discussions tend to meander a lot and lose focus. For that reason,
>> please write me email about joining a mailing list dedicated to
>> discussions on object lifetime - a study group. The outcome should be
>> a DIP that moves forward safe lifetime management for D.
>
> +1, but creating a dedicated mailing list feels exclusive to me again.

I don't plan to be selective about admission. Whoever wants to join can join. All we need is to not make the discussion part of the general forum entertainment.

> I
> would at least like to read what they are talking about once a while.
> Maybe mark the threads "experts only" or something.

So I'll add you to the list then.


Andrei

On 10/27/15 5:01 AM, Dicebot wrote:
> Invite-only mail list with archive being published via forum.dlang.org
> interface (read-only) sounds like a best match to be.

Ah, interesting. Vladimir, do you think you could rig such a list? -- Andrei

On Tuesday, 27 October 2015 at 09:49:48 UTC, Andrei Alexandrescu wrote:
> On 10/27/15 5:01 AM, Dicebot wrote:
>> Invite-only mail list with archive being published via forum.dlang.org
>> interface (read-only) sounds like a best match to be.
>
> Ah, interesting. Vladimir, do you think you could rig such a list? -- Andrei

I can add it to forum.dlang.org after Brad sets up the actual mailing list. I don't host any mailing lists myself.

Although I could in theory begin hosting mailing lists, the one time I tried it was very unpleasant and non-trivial and seems like a disproportionate amount of effort for this project.

What's wrong with a thread on the current MLs? I might be wrong but it seems that the amount of discussion on how to discuss the matter is already disproportionate to the amount of discussion on the matter itself. Besides, we have too many dead MLs already (dmd-concurrency? Digitalmars-d-dtl?)