Re: josm orphaned, or why are we packaging

[Przemek Klosowski]

On 03/22/2018 07:55 AM, Ralf Corsepius wrote:
>  On 03/22/2018 11:32 AM, Matěj Cepl wrote:
> > On 2018-03-22, 06:51 GMT, Till Maas wrote:
> >> I orphaned josm 
> >>
> (https://na01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fsrc.fedo...),
>
> >> the
> >> java openstreetmap editor on request by the original
> >> maintainer. Please adopt it. It needs to be updated regularly
> >> to follow the current openstreetmap guidelines, currently it
> >> is outdated (also in EPEL).
> >
> > My experience with JOSM is that this is probably one of the
> > programs which are better not to be packaged (other example: vim
> > plugins). Java Web Start on
> >
> https://na01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fjosm.ope...
>
> > works just fine
> > and I don't see the reason why we should bother with packaging
> > it ourselves.
>  The reasons for packaging something up as rpm is security and system 
>  integrety/consistency, i.e. not to endanger installations from the 
>  (windowish) mindset of "unpackaged" third party stuff.
>
>  What you say, basically means you are questioning and deny the 
>  usefulness of packaging as a whole. The key feature which has made 
>  Linux distros great and superior to Windows.  
I fundamentally agree with Ralph, but
I also wanted to point out that 
the situation is more complicated than it used to be. The executable 
content is available a spectrum of ways, from most heavyweight to most 
lightweight

 1. ELF binaries
 2. binary run-time loadable libraries
 3. script files for scripting language environments (java programs
    could be arguably placed here)
 4. scripting language libraries
 5. java applets running in the browser
 6. javascript running in the browser

Clearly, we want to package 1. and 2., and we aren't going to start 
packaging 6; there's a big discussion as to what's the right approach 
for 3 and 4 (npm, conda, cargo, etc).

My point is that we have to decide where we put the line; what Ralph is 
saying, and I agree, is we want to push it as far down as 
practical---but there always be some executable content that doesn't 
make sense to be packaged. We just need to decide why we're packaging 
and what are the tradeoffs.

One way of looking at it is that packaging provides an assurance that 
the software we're running is the software we think you're running, as 
opposed to downloading random binaries and scripts from the internet 
(curl | /bin/sh). In this way of thinking, software downloaded from 
secure (TLS/https) connections to trusted sites could  be considered as 
good as packaged---we're doing it to javascript so why not java and 
other things. The .jnlp file that provides JOSM is essentially an XML 
file which starts the java machinery running the OSM-provided java 
application--I can see how people could argue that it's no different 
from maps.google.com starting a javascript mapping application in your 
browser.

Another criterion could be whether the software installs itself on your 
system, or is transient, putting the line pretty firmly between 4. and 
5. That would push the JOSM into the packaged category. This approach 
addresses the bitrot aspect---the fact that unpackaged software has a 
tendency to fall behind and end up with bugs and exploits.

I think that we should reflect on such fundamental requirements because 
they are relevant to how the modularity and container situation develops.