[j-nsp] EX4200 9.4R2.9 process crash on previously valid config

[Richard A Steenbergen]

On Fri, Apr 03, 2009 at 09:57:36AM +0800, Mark Tinka wrote:
> I realize that Juniper have to make a new release every 
> quarter, and in most (if not all) cases, there is some new 
> feature which has the potential to cause "badness" to 
> systems that are already working fine.
> 
> I think what I'd like to see is a fork of the code base 
> which consolidates all new features up until that point, and 
> then enters into a maintenance state - only bug fixes, with 
> no new features.
> 
> Other iterations of the code would continue to come out with 
> new features every quarter.

In theory thats what should be happening today. Each minor version bump 
(i.e. 9.3->9.4) is effectively a new code base which is developed 
independantly from the previous version. A revision bump (R1->R2) should 
be entirely bug fixes, no new features should be added, and all feature 
development is occuring in future versions of code (i.e. 9.5 etc) which 
have not been released yet.

But somehow they're managing to break a whole heck of a lot of stuff
with every version, not just things related to the "new features", and
many times between R version bumps as well. I really don't know enough
about Juniper's development process to say anything intelligent about
what the problem is, but it is a serious problem for those of us trying
to run networks with it. My personal suspicion is a diluted focus on
software for a stable core router, in favor of trying to stuff in
features for firewalls and every other weird enterprise box that Juniper
has branched off into. I'd also guess that the "one single software
image" marketing line is now more of a liability than a benefit.

I'll stop short of trying to tell them how to do their jobs (since again
I really have no clue about most of the inner workings :P), but I think
they could absolutely use a kick in the ass from customers to remind
them that software stability needs to be a high priority again. Some
attention to feature requests and customer feedback purely because it is
the right thing to do rather than because it is a large account who
requested it would also be a welcome throwback.

And for the record, we all love Juniper and JUNOS (even those of us who
gripe about the problems :P), we just don't want to see it continue to 
become a victim of its own success. If I want bloated buggy crap that is 
always a constant source of grief, I already have IOS.

-- 
Richard A Steenbergen <ras at e-gerbil.net>       http://www.e-gerbil.net/ras
GPG Key ID: 0xF8B12CBC (7535 7F59 8204 ED1F CC1C 53AF 4C41 5ECA F8B1 2CBC)