On Wednesday, June 24, 2015 at 9:03:15 AM UTC-5, miguel wrote:
>
> On Wed, Jun 24, 2015 at 12:26 AM, Jason Slagle <rais...@tacorp.net 
> <javascript:>> wrote: 
>
 

Open source users will always rely on a third party for security 
> patches anyway. Every single vendor, named CentOS, Debian and Ubuntu 
> drops the ball from time to time. Go and look around about the OpenSSL 
> problems Debian had, and the long period CentOS users were completely 
> in the dark for security patches. 
>
>

"Always" is a bit strong, but yes, it is typical these days to rely on 
third-party packagers for updates of all kinds, including security 
patches.  No one is ever locked in to that, however.  An AIO package does 
make it harder to apply upstream patches, though, and as the number of 
upstreams contributing to a single package increases, either the frequency 
of package updates must soar or the (average) lag between patch release and 
package update must rise.

Moreover, the concern about PL's commitment over time should not be 
discounted.  PL implements a pretty aggressive development and release 
cycle, and having many simultaneously supported versions will have an 
additional combinatorial effect on them.  Historically, they have supported 
their software for relatively short periods as many enterprises account 
things, and once a product goes out of support, you can no longer expect 
updates from PL for *any* of the pieces.
 

> > A related concern comes with companies with infosec departments that 
> have 
> > to bless things.  I get Ruby 2.1.0 blessed, but then the bundled ruby 
> gets 
> > updated to 2.1.1.  Now there are a lot more compliance hoops to jump 
> > through. 
> > 
>
> One of the tradeoffs of using the AIO is to be able to move faster, to 
> get more done, instead of having to be limited by the least common 
> denominator.



I'm sorry, I don't follow that logic at all.  I can see the vague outlines 
of an argument along those lines about AIO on-disk layout and vendored 
components, but that's not what we're talking about.

 

> Companies with infosec departments that have to bless 
> things, IMHO, that is their problem. Don't want to fight the 
> bureaucracy? Buy the support and outsource the responsibility. 
>
>

That's a rather cavalier attitude, not to mention missing the point.  Even 
companies that pay for support may still have a requirement for the 
software to be blessed.  That you personally have low regard for such 
requirements does not make them irrelevant to the discussion.

 

> > In the end, a lot of it comes down to it “not being the unix way”.  I 
> have 
> > many of the same arguments and dislikes against systemd.  I have no 
> issue 
> > with the AIO installer, and in fact might use it on some older 
> > centos/rhel5 hosts where getting modern ruby is hard.  My heartburn 
> comes 
> > from it being the only REAL way to install these packages starting with 
> > version 4.  I’d much prefer you also support a more traditional 
> > metapackage approach for the operating systems that support it. 
> > 
>
> The Unix way? Come on. Lets get rid of packaging then. I'm 34 and the 
> more senior people I've spoken told me that back in the old days they 
> downloaded the source code and compile it, on every machine, with that 
> nostalgia look in the eyes.
>


Yes, I have no intention of going there.  Packaging is one of the 
significant advances of the last 30 years or so.  That's why there are so 
many packaging systems.  As for "the UNIX way", UNIX architecture and 
tradition revolves around providing many small(ish), independent components 
that can be combined in useful -- and often unforeseen -- ways.  To some 
extent, AIO packaging indeed does run against that, and to that extent I 
agree with Jason.

 

>
> I think it is counter productive to stick with the metapackage 
> approach when there is going be a burden to maintain some operating 
> systems in one way, and some in another way. 
>


That is certainly a consideration for *PuppetLabs*, and it seems clear that 
they have taken it to heart.  I am by no means convinced, however, that it 
would be harder to following the metapackage approach for those packaging 
systems that support it than not to do so.  Any way around you need 
different packages for different systems.  No matter what, it is not a 
consideration that matters *to me*, nor, I suspect, to most other users.


John

-- 
You received this message because you are subscribed to the Google Groups 
"Puppet Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to puppet-users+unsubscr...@googlegroups.com.
To view this discussion on the web visit 
https://groups.google.com/d/msgid/puppet-users/16be4a74-57db-4e0e-a2c4-2657a9922e13%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Reply via email to