Are We Dead Yet?

[PascalDragon]

I suppose that one day we'll need a RISC-V port ....

RISC-V support (both 32- and 64-bit) is already available in trunk and will be part of the next major release.

[dieselnutjob]

I suppose that one day we'll need a RISC-V port ....

RISC-V support (both 32- and 64-bit) is already available in trunk and will be part of the next major release.

Damn. That means I will just have to buy more toys from Pine64.  I already seem to have most of their products.  They are just too tempting.

[LV]

10000000 downloads of Lazarus. Congratulations

[SymbolicFrank]

Very nice!

[lazer]

I justcounted the posts on the fpc-devel mailing list:

2018: 414
2019: 766
2020: 457
2021: 341
2022: 332
2023: 200

This gives  a linear regression of  posts =  -71.09*year + 144047, which predicts that in three years (i.e. 2026.26), there will be no more posts at all on the  mailing list.   It is as good as dead. Not even Lazarus could resurrect like that

You have all the statistical skills of a climatologist.

Below is your data and regression fit with a more useful scale.
Do you think your proposed linear model is a good fit to the data. Does it capture the key features?
What was the uncertainty of the fitted coefficients?   (please don't tell me you hit the "fit trend" button in Excel)

gnuplot:

Code: Pascal  [Select][+][-]

1. Final set of parameters            Asymptotic Standard Error
2. =======================            ==========================
3. m               = -71.0857         +/- 36.85        (51.84%)
4. c               = 144047           +/- 7.446e+04    (51.69%)
5.  

What is the range of the zero crossing point with a 50% change in slope?

When doing a regression you need to explain why you choose a particular model for the data. Just saying the word "trend" is NOT a valid reason.

"which predicts that "

No the regression of an arbitrary fn does not "predict" anything. It is you which is attempting a prediction based on the extrapolation of a questionable model for the data.

In 2007 a negative  quadratic  made a reasonable fit to the Arctic ice sea extent since 1978, leading to Noble "Peace" prize winners like Al Gore and activist-scientists like Cambridge Prof John Wadhams to predict ice free summers as early as 2013.

That choice of regression model was based on unfounded, a priori assumptions about the causes of the melting. The summer minimum extent in 2023, 2024 was indistinguishable from what it was in 2007.  No net change in 14 years. Happily, our "canary in the coal mine" is still chirping.

It is important to chose a suitable model for the data to avoid imputing erroneous characteristics on to the data.
"Trend" is not a predictor of anything apart from a lack of understanding of statistics.

We now have the potential for two more data points , how does that look?
How has the sad demise of the founding father of fpc affected that curve?

[dbannon]

Lies, damn lies and statistics ?

Your result, clearly with some expertise and effort produced pretty much the same result as Benny's light hearted guess. Yours implies a little under 200 for 2024. In fact its 344.

Oh, the danger of statistics!  To follow on from your example, confusing weather (ie annual (arctic?) ice) with climate (say, ten year averages https://earth.gsfc.nasa.gov/cryo/data/current-state-sea-ice-cover). And, the oh too human tendency to select our data to suit our preconceived model. And assume one bad forecast nullifies hundreds of proven good ones ?

Anyway, Benny's data ?  Maybe commit count would be better ?  Or, perhaps, better still, a count of commits that resolve a bug report ? Not a lot of recent data if we were to work on time between releases 

Why don't you try to plot that ?
324  49 months and counting
322 Aug 2021 14 months
320 Jun 2020  31 months
304 Nov 2017  9 months
302 Feb 2017  14 months
300 Nov 2015

(Hey, we missed 3.2.2's 4th year birthday !)

Davo

edit: corrected spelling, thanks Zvoni, I have replaced the keyboard in my laptop once already !

[Zvoni]

Oh, the danger of statics ! 

Don't you mean "Statistics"?

Well, to quote Winston Churchill:
"I don't trust any statistics i haven't forged myself"

[Khrys]

How has the sad demise of the founding father of fpc affected that curve?

The founding father of Pascal recently passed away, but the founding father of FPC seems quite alive to me.
Not surprised to see a necroposter hung up on dead people...

[lazer]

The founding father of Pascal recently passed away, but the founding father of FPC seems quite alive to me.

Ah, phew! thank you for the correction. I was shocked to read that news and must have misread it. It was indeed Niklaus Wirth who passed away.

My sincere apologies to FPK for the careless error.  I wish him a long and healty life.

[lazer]

Your result, clearly with some expertise and effort produced pretty much the same result as Benny's light hearted guess. Yours implies a little under 200 for 2024. In fact its 344.

NO. My analysis does not "imply" anything. Now it is you who is implying not Beni. It's amazing how many people think that you can fit a straight line to any data and that this automatically has predictive ability about the future.


 I get a similar regression result by applying the same ill-founded model. My whole point is that this implies nothing.
Firstly you need to apply +/-50% to the slope and give the zero intercept of that interval, even if the model made sense which it does not.

And, the oh too human tendency to select our data to suit our preconceived model. And assume one bad forecast nullifies hundreds of proven good ones ?

When was there ever a good one?  Climate models failed to predict the rapid decline from 1997 to 2007. So they arbitrarily frigged the fudge factors to get a better match. Then the ice stopped disappearing and the models were far too pessimistic. Now they make handwaving claims about : not ice free before 2040 or 2050. They are still totally ignorant of the underlying physics driving the system but 2050 is far enough off we'll have all forgotten their silly "predictions" by that time they can be tested against factual data.

The summer arctic ice minimum was the most dramatically changing metric, so it was presented as "the canary in the coal" of global warming.  Since the record minimum of 2012 we don't hear a peep about it. As you say "select our data to suit our preconceived model.  " . Exactly. Then when it doesn't fit any more ... hey, look up there !

 Maybe commit count would be better ? 

Yes, I agree.

Why don't you try to plot that ?
324  49 months and counting
322 Aug 2021 14 months
320 Jun 2020  31 months
304 Nov 2017  9 months
302 Feb 2017  14 months
300 Nov 2015

Not much point with the sparcity of recent data. If you can pull out monthly or at least annual commit count that would be more interesting.

[ccrause]

Anyway, Benny's data ?  Maybe commit count would be better ?  Or, perhaps, better still, a count of commits that resolve a bug report ?

I agree with the sentiment about looking at one metric only.  From my observations of the chatter on fpc-devel and Gitlab issues and merge requests, I suspect that more of the specific discussions have moved from the mailing list to the Gitlab platform.

[zeljko]

https://hackaday.com/2025/09/02/the-case-for-pascal-55-years-on/