Re: The (dark) future of Java on desktop

[Dirk Stöcker]

On Thu, 8 Mar 2018, Richard wrote:


On Thu, Mar 08, 2018 at 08:36:21AM +0100, Frederik Ramm wrote:


You could sit down today and re-implement everything in, say, C++, and
it would be relatively straightforward, and while the result would not
share any of JOSM's codebase, it would still encapsulate all the
experience and brainpower that has flown into JOSM development over the
years.


true in principle but you would need a protable GUI that doesn't suck or
you end up programming for at least 3 platforms with 3 sets of bugs,
3 sets of dependencies etc.


Reimplementing an existing software like JOSM which has an estimated cost 
of more than hundred development years (https://www.openhub.net/p/josm) in 
another language in an non-profit OS application is doomed to fail in my 
eyes. The motivation for a programmer to take an existing software and 
reimplement everything again is low. For a very long time you will not 
have something which is usable and inbetween you have tasks to do, but no 
positive feedback. That may work when the people are paid for it, but not 
when programmers need to be motivated. I'd consider people beeing motived 
by such a task very strange. :-)


Rather than that if JOSM really dies some of the better ideas of it will 
be taken and implemented in existing or new software (which BTW is already 
happening, e.g. osmosis taking the Validator MapCSS or many other things).


If there is a way to automatically convert the code and start with a 
working base, then the situation is different...


But I also don't think this is necessary (ATM).

Ciao
--
http://www.dstoecker.eu/ (PGP key available)

Re: The (dark) future of Java on desktop

[Michael Zangl]

I don't think it's worth speculating now.

(1) The linked document mostly focuses on Java web start (which is just
a convenience for some users - but we can ship JOSM otherwise, e.g.
using a launcher like many java programs do) and other non-JOSM
technologies (we have only one class depending on JavaFX)

(2) For Swing/AWT (that's what we use), I could not find any hint that
it will be dropped. Instead, Oracle is stating that it will be supported
at least until 2026 - that's 8 years. Finding a programming framework
that will be supported in 8 years is very difficult. Even web browsers
change faster than this and many web applications that worked 5 years
ago are already broken.
(They mentioned the java 11 release explicitly because it is a LTS release)

Don't worry, we will be fine with java for now ;-).

And if one day we won't, computers will be so powerful that we can
transpile everything to javascript :D

Michael

Am 08.03.2018 um 23:54 schrieb Wiktor Niesiobedzki:
> My reading of this Oracle post is that is to actually change the way you
> ship the applications. Instead of relying on JRE installation on client
> station - ship your code bundled with JRE as jlink does (and take care
> about all the updates yourself).
> 
> Anyway I guess that we can assume that number of end-user installations of
> JRE will be shrinking, so shipping JRE together with your application might
> be already a good idea. We should watch what Eclipse will do about it (and
> all the commercial tooling based on Eclipse). My guess is that they will
> not give up on it so easily.
> 
> It means then that we need to cover all platforms in our build system. This
> would be the case also whatever programming language we will take.
> 
> If JOSM were to abandon Java as a language maybe we should think about
> extending/repackaging/repurposing QGis? I guess that probably there were
> such ideas in the past?
> 
> Cheers,
> 
> Wiktor
> 
> 2018-03-08 19:07 GMT+01:00 Vincent Privat :
> 
>> WebStart is going away. It is the only part of Java that isn't open source
>> and they explicitely stated they won't open source it:
>> https://twitter.com/DonaldOJDK/status/971492781616136194
>>
>> So at least starting from September we'll have to make the WebStart link
>> less prominent as it won't work anymore for Windows and macOS users having
>> their Java up-to-date. It will work natively only on Linux, where openjdk
>> package includes the retro-engineered free version of WebStart
>> (netx/icedtea-web):
>> https://icedtea.classpath.org/wiki/IcedTea-Web
>>
>> I don't know if the few (single?) people behind IcedTea-Web will have the
>> desire to maintain it after Oracle drops it from JDK. I don't know either
>> if icedtea-web requires jar signing: maybe we will be able to drop the
>> requirement to sign josm.jar, thus asking to Frederik to pay for the
>> certificates ;)
>>
>> JavaFX will be given to someone else soon. Maybe the Eclipse Foundation,
>> like Java EE which has be transferred to Eclipse, without the permission to
>> call it Java EE anymore. Or maybe the Apache Foundation, where they already
>> made OpenOffice and Hudson die there (given that they have been
>> successfully forked as LibreOffice and Jenkins).
>>
>> AWT and Swing will still be here in Java 11. But the fact they mention it
>> explicitely today probably means they have plans to remove it as soon as
>> Java 12 development starts (in 6 months). I have no idea if the new project
>> will create enough traction to have enough contributors (volunteers or pais
>> staff from other companies), we'll see. Swing is still used a lot in the
>> industry. At least we should be able to fix Swing bugs ourselves when we
>> find ones.
>>
>> Concerning JOSM it means we will probably have to ship AWT, Swing and
>> JavaFX in josm.jar. In JDK9 the desktop module (AWT+Swing) weights 13Mb,
>> the various JavaFX modules 30Mb. JOSM jar is only 12Mb today.
>>
>> Cheers,
>> Vincent
>>
>> 2018-03-08 11:19 GMT+01:00 Dirk Stöcker :
>>
>>> On Thu, 8 Mar 2018, Dirk Stöcker wrote:
>>>
>>> [nothing]
>>>
>>> Sorry, operator error :-)
>>>
>>>
>>> Ciao
>>> --
>>> http://www.dstoecker.eu/ (PGP key available)
>>>
>>