I have made a small project that uses 2N3819
I used the one comes with KiCAD
It says
1 = Drain
2 = Gate
3 = Sources
confirmed by the PDF link is into the symbol as well
But all the other datasheets on the ned are telling that the 2N3819 inthe reality is
1 = Source
2 = Gate
3 = Drain
So 2 and 3 are inverted in KiCAD
Interfet
Vishay
Onsemi
If you have dueling data sheets then how can you say one is wrong? Obviously this is manufacturer specific even if it is only ONE that is at odds. Been awhile since I’ve done a board but I remember choosing the symbol by the datasheet of the component I had.
I guess whoever did the symbol used the Central version. I guess this is why the database approach is gaining steam?
Because it is the only one wrong. All the other ones are according to each other.
EDIT: I have several schematics from other manufacturers of electronic devices they are indicating 3 = Drain :-/
It isn’t wrong if their component is manufactured that way is it? An outlier, but not wrong. Why it is different is open for speculation but unless you test one from Central and find it isn’t according to the datasheet, then no, it isn’t wrong. It is different.
Now, would it be better to change the library symbol to a more industry compliant version? Probably.
ok however I used that JFET standard (the only one 2N3819) in KiCAD and of course the JFET I used is the standard pin out; The one in Kicad isn’t.
What’s a “JFET standard pin out”? Never heard of one.
And with devices this ancient, it’s not unusual that different manufacturers had different pin out.
To avoid such situations, use the generic symbols in the “Device” library and select the one with the correct pin out according to your actual data sheet.
But now this problem has been solved! Every new device is made by only one manufacturer; no second sourcing! 
Just turn the TO-92 case the wrong way at assembly time et voila two wrongs make a right. You’re lucky it isn’t the EBC vs ECB pinout which requires more fancy legwork. 
On the Z axis! Rotating 180 degrees on X or Y axis will result in all 3 leads being inverted. Really inverted, as in upside down. The leadwires would end poking up in the air and not passing through your through hole pads. There is an advantage to this, though: Minimum power consumption.
Disadvantages: The transistor is likely to fall off the board and be lost. The design would experience reduced functionality, assuming that the FET was needed for proper functionality.
Do you mean they make a left?
But seriously: Pin numbering (mostly) and pinout non-conformities have been a recurring theme in a number of threads on this forum. This seems particularly true for SOT23 and maybe some D-paks and D2-paks for example. And forget about the exact spelling of D-Pak, with or without a “c” or a hyphen. I think that DIPs and SOICs (for example) mostly escape this problem.
The only good solution is be careful and check the datasheet. Hopefully the manufacturer has not made an error on the datasheet…
It’s a 2N3819 did I mentioned? Even if I didn’t tell: its blatant from the datasheets.
Ah this occurred already indeed 
Yes I watched the datasheet from the supplier. While the datasheet used to make the component library was different.
I discovered this AFTER 5 PCBs were manufactured and delivered with welded components … being for microphone, it’s better each pin is correctly connected.
ON the datasheet of that particular JFET is not written is flippable :-/
Many, but not all, jfets are symmetrical source - drain, so flipping would have no practical effect.
…; yes but in these cases it’s reported on the datasheet (like the 2SK170). I didn’t see it on the 2N3819.
Back in the mists of time, I had a jfet databook that actually showed the (small number of) die from which the various jfet types were selected during packaging and test.
I am fairly sure that the 2N3819 was one of the symmetrical parts. Identical Vdg and Vgs ratings are a hint.
It was Nat Semi. This book was important back then in the 70s…
Trust me when I say trust no-one. 
But seriously feel free to submit a new symbol 2N3819_SGD and the minority one can be renamed 2N3919_DGS.
Tell me where to send new symbol etc and I will prepare it.
Gitlab, KiCad library, merge request. But don’t you need some anti-gitamine first? Or was that the other guy?
Yes from the Die it seems very symmetric indeed
