[j-nsp] Stable Junos

[Richard A Steenbergen]

On Wed, Sep 01, 2010 at 01:55:49AM +0800, Mark Tinka wrote:
> 
> I wish 10.5 were coming out tomorrow, but for now, 10.2R2.11 has been 
> behaving with the little we're doing. I'm tired of chasing JUNOS for 
> the past 2 years, so I wouldn't mind continuing with 10.2R2.11 until 
> 10.5R4 or 10.6R4 comes out (and works). Of course, that's hoping we 
> don't run into anything major with 10.2R2.11 :-\.

It's like playing JUNOS Hero, you don't ever catch the dragon. :)

Personally I'd like them to support JUNOS images for a bit longer. By 
the time a particular branch is actually stable enough to use without 
major issues (R4) it's already EOL, which means they refuse to build 
fixes for any newly discovered bugs after that. Infact for the last 
several releases they've actually released R4 *AFTER* the EOL date, so 
you're guaranteed that anything you find wrong in R4 will never get 
fixed. This forces you to chase the next newer version of code to fix 
those last few issues, where for every 1 thing they fix they break 2 new 
things, then rinse and repeat forever.

The extended support branches never quite seem to line up with anything 
we need either (I think they're picked solely based on what satisfies a 
few large accounts with some T1600s :P). For example, the extended 
support branches right now are 10.0 and 10.4. 10.0 is worthless to 
anyone who needs Trio card support, 10.1 reaches end of engineering 
in 2.5 months, this pretty much guarantees we'll be chasing 10.1, 10.2, 
10.3, etc, in the above cycle. By the time 10.4 actually gets stable 
enough to use in widespread deploys, which is at least a year from 
today, where will be some other reason why we're forced into something 
newer. There always is. :)

-- 
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)