D1 to be discontinued on December 31, 2012

[Andrei Alexandrescu]

On 12/14/11 2:30 AM, Don wrote:
> On 14.12.2011 05:37, Andrei Alexandrescu wrote:
>> There is no abandonment. Also, where is that 50/50 estimate from? Just
>> curious.
>
> The D2 community is definitely bigger than the D1 community. But how
> much more?

I presume it's quite a bit larger. But then both are small, and we're 
interested in the potential and the rate of adoption.

> It's hard to be sure, but the Tango users used to be 75% of the
> community, based on a few polls that were held, but they never had much
> representation on the ng. I guess between half and 2/3 are gone now.

That seems a reasonable assessment. Possibly even more left.

> I don't think the entire D community is as big as it was back then
> (based on number of public repositories).

That is also entirely possible.

> Additionally, the number of contributors, and level of activity, in
> Tango, was higher than Phobos has ever had.

Agreed.

But this is all missing the mark - why would we cry over spilled milk. 
The point is assessing the state of affairs the minute before the 
announcement. How active was Tango? How active were the Tango forums? 
Where were other forums of the D1 community? In this day and age, I'd be 
hard-pressed to think of an active programming language community that 
has no online presence whatsoever.

To add to that, there was no trickle of bug reports or pull requests for 
D1, although clearly D1 does have its bugs and issues. I haven't 
followed Tango closely, but if D1 had a large active community, Tango 
would receive a lot of attention from its users as it's the de facto 
standard library for D1. Yet the last post 
(http://www.dsource.org/projects/tango/forums/topic/903) dates from 
March 30. The intervals between changes to the trunk 
(http://www.dsource.org/projects/tango/changeset/5691/) are measured in 
months.

I hope you'll agree that one would be hard-pressed to infer from the 
evidence there is that there's a large, active, and thriving D1 community.

>>>> It also diffused our focus, delayed us to an ever-increasing extent,
>>>
>>> What on earth gives you that idea? The only resources involved are some
>>> fraction of Walter's time, which is obviously an important resource, but
>>> nobody other than Walter is affected.
>>
>> Walter told me so, and Walter is the single most important bottleneck.
>
> So "us" means Walter?

A delay originated by Walter delays the entire community. So by "us" I 
meant the community.

>>> You're seriously misrepresenting
>>> the situation.
>>
>> What is an accurate representation of the situation, and what evidence
>> is there to back that up?
>
> This is an issue of Walter's time, and nothing more.

Walter said that he'd rather not spend that time anymore.

> I think the
> evidence for that is clear -- who else is spending time on D1?

As I mentioned, the larger issue is the message we're sending out. We 
can't serve two masters, and we need to focus and specialize our brand. 
Please trust me on this; in only a few short months I'm confident you 
will agree this will have been a good move.

> Important thing to notice: GDC and LDC also have D1 compilers. Does this
> decision apply to them?

No.

>>> I spend some time on fixing D1 bugs, but that won't change, see below.
>>>
>>>> and
>>>> sent the wrong message out that we're lacking confidence of what our
>>>> core thrust is, so we're trying to sort of please everyone. ("Here's
>>>> our
>>>> flagship language! If you don't like it, well, we have another one.")
>>>
>>> This is a silly and offensive statement. Most languages are in this
>>> situation. Look at Python2 vs Python3, Perl6 vs Perl5.
>>
>> I wonder who would be offended by that.
>
> Other than me? Well, I think the Python guys would be offended if you
> said it of them.
>
>>>> 2. The deadline is more than a year away. This is a long time, enough
>>>> for us to make D2 compelling, and also for interested people to
>>>> migrate.
>>>
>>> No, it's an exceedingly short timeframe.
>>
>> Other vendors give similar time scale for much larger migrations.
>
> Only when the migration is to a mature product.

Agreed. All things considered, one year seems to me a resonable time 
frame. It's reasonable of you to disagree. Time will tell whom of us was 
right.

>>> What gives you the idea that
>>> nobody is using D1?
>>
>> There is not one sentence in my message claiming nobody is using D1.
>
> Ok, I can't make sense of it otherwise.
>
>> Don, it is you who is misrepresenting the situation, and repeatedly. I
>> understand you find this frustrating, but please, let's have a
>> constructive dialog.
>>
>>> Have you thought about what would convince them to
>>> switch to D2, and what would be required for them to do it?
>>
>> I think the most important aspect for them would be completion of
>> Tango's port to D2. The recent progress in the matter is encouraging. If
>> the D1 community is sizable, resourceful, and interested, I believe that
>> to be within the realm of possibility.
>
> At the present time, do they actually desire to move to D2? If not, why
> not? And the big one: are they confident that D2 is sufficiently stable?

I don't know what they think. Based on the observations I outlined 
above, I can hypothesize (a) the interest to move to D2 is scant; (b) 
the main reason is Tango is not working with D2; and (c) I don't know 
and I don't think that's the gating factor (Tango is). I'll venture the 
opinion that the D1-like subset of D2 is as stable than D1.

>>> I suspect you don't know much about the D1 community. (Note that only a
>>> small fraction of D users have ever used the newsgroup, and it's mostly
>>> people with an interest in language design. They are not
>>> representative).
>>>
>>> I can already say with certainty that I will still be using D1 in 2013.
>>
>> That's great. The decision does not make it impossible or even
>> particularly difficult for D1 users to continue using D1. Since there
>> were near zero bug reports on D1,
>
> There are 400 open bugzilla D1 + common D1/D2 compiler bugs in Bugzilla,
> including nearly all of the heavily voted bugs. D1 is very much more
> stable than D2 (D2 has 800 open bugs), so there are far fewer bug reports.

More bug reports does not imply less stability because number of users 
also affects the numbers. Also, there is heavy community activity to fix 
D2 bugs but not D1 bugs. It's reasonable to assume there aren't many 
people waiting for some D1 bug to be fixed. They either found a 
workaround or left the language.

> If you mean, there were no bug reports on the beta, that's not
> suprising, there are no D1 Phobos changes, which is where regressions
> mostly show up.

I wasn't referring to this beta as much as the overall pattern of 
reporting D1 bugs, discussing them, and contributing fixes to them.

> they can be assumed to be content with
>> the quality of the compiler. Really I don't see the gist of the
>> complaint. This is not abandonment.
>
> You, Andrei, are personally making a decision which affects the entire
> community. I'm actually a bit shocked that you've done this.

Walter and I made this decision together. This is not a decision of the 
community because, again, people like to postpone, keep options open, 
and be generous with others' time. To give a counter-example, I asked 
for community input on the new website although I was advised against 
it. Indeed I've been racked over coals, but the ultimate result is a 
much better website than it would have otherwise been.

If I knew of your response, we would have made sure to pass it by you, 
and I apologize we didn't. (I assume the entire core team would be on 
board with the decision, and aside from you, my perception is that it 
is.) But if you have already enumerated the arguments you would have 
made prior to making the decision, I think you'd agree they wouldn't 
have been compelling.

> Especially since there was a big public discussion on the D internals
> newsgroup (Nov 10) about reducing the pressure of Walter's time.
> ---------
> Kenji:
>  > Today only Walter improvements D1 branch. Almost dmd pulls only
>  > consider D2 branch. Should we add D1 patch at the same time?
>
> Walter:
> Merging with D1 hasn't been too difficult; I use a program called "meld"
> which makes it a snap.
> ....
>
> It usually takes a couple hours to merge a patch, if things go smoothly.
> Most of that is running the test suite. Some more time is spent updating
> the changelog and bugzilla.
> ----
> So, didn't sound then as though D1 was a big issue. Yet a month later
> you announce you've made a private decision about axing D1.
> Feels exactly like a military coup.

Well I am as surprised about your surprise as you are about the fact 
itself. About this, I repeat that Walter's time is one component, but 
not the major one. We can assume it costs him zero time; still the 
decision to announce ceasing D1 support in one year is a good one. 
Please trust me on this one, and may D1's blood be on me if I made a 
mistake.


Andrei