I said that download numbers are not reliable as indicator for eclipse continued use.
Get numbers that indicate *usage* - I gave JBoss Tools usage numbers as indicators as that is a tool set that historically been growing at same/similar speed as eclipse adoption/usage and it last year for the *first* time saw a decline - and still it is in top 7 of eclipse marketplace. WildWeb is doing great but wildweb being great isn't indicator for whether overall usage are dropping.
Thus instead of going into a battle of which distro is better than something else; go fetch the http post 200 requests logs that are to p2 updatesites everytime a *actual running* eclipse install do daily/weekly update check and use that number as indicator.
Eclipse usage *is* dropping/going downwards - its very clear to see in combined surveys and general conversation with people at general (non-eclipse specific) specific events. I'm though still convinced eclipse is way more used than popularity surveys indicates; but to battle that and not be blindsided by ones own biases (mine included) go look at *usage* numbers (i.e. updatesite pings/checks) not downloads nor indidivual plugin downloads.
Please :)
On Thu, Feb 6, 2020 at 9:56 AM Aleksandar Kurtakov <akurtako@xxxxxxxxxx> wrote:
I fully agree snyk report about massive drop is not true for the wide industry but using download numbers ain't going to give a true picture either.
But looking at jboss Tools usage stats there are a decline the last year for the first time in almost 10 years so things are changing.
IMHO this is because JBoss Tools still ships the old WTP-based JS/HTML/CSS/etc. editors and their state is not good enough thus WildWebDeveloper is eating their share.
Eclipse doesn't have similar usage tracking but there are P2 update http requests made.
So Look at P2 update stats hit instead of downloads. That will be a much better unbiased source.
Note that the numbers have been unreliably low for an unknown
period of time since before Denis fixed the problem with rolling
up the daily stats (2019-01-09) into cumulative stats:
Goodness knows which 2000 responses were measured by this survey
nor how representative that is of the entire Java ecosystem. I'm
doubtful it's in any way representative of the broader reality.
For example, based on this report, apparently close to 50% of the
Java developers pay for and use Intellij IDEA. As such, given the
huge Java market in which apparently 50% of those developers are
paying some $ for Intellij, they must be making a massive fortune
from the Java ecosystem. How likely is that in reality? It would
certainly help them invest far more heavily than does Eclipse's
freetarded ecosystem.
In the end, I don't see much value is looking at zipped update
sites. Who actually does that?
Regards,
Ed
On 06.02.2020 09:57, Ed Willink wrote:
Hi Mickael
I'm not at all sure how well the download stats relate to
reality...
I was looking at the download stats a couple of days ago to try
to determine how dead MoDisco is...
It shows small values, almost like a background level, but why
did 10/26 of the 1.5.0.zip downloads occur on Xmas day?
Why is 1.5.0M3.zip still being downloaded - albeit trivial
number of times.
Looking at a more active project ATL-Update:
why only 7 download in 35 days of 2020 compared to 5711 in
2019?
Changing to total per file, we see the top 20 entries with
between 103 and 193 downloads. OK a very even spread ?? good,
but most of the entries are milestone builds. Why are so many
'users' sticking with an even spread of milestone builds? Surely
this is an indication that the 'users' are 'spiders' or 'mirror
polling' rather than true users.
Changing to EMF-Update we see results that might be big enough
to escape the 'background' noise. The top 5 entries are the most
recent 5 releases 563 to 185 downloads. Plausible, but the sixth
entry is 2.6.zip; a 10 year old release; why is this still so
ridiculously popular? It doesn't correlate with the last Eclipse
3.x release; It is a download from the latest organisation of
EMF downloads that did not exist until 8 years after the release
was superseded; more traditional downloads were emf-xsd-update.
----
My examples above are all on smaller numbers and so the
associated issues may be swamped by the larger numbers for the
SDK, but I think there are sufficient uncertainties to have
doubt as to whether the numbers tell us what we think they do
about 'users' rather than 'bots'.