Varnish-3.0.0: All this bits are yours...

[Poul-Henning Kamp]

So we've gone and done it:  You should get this message right after
Tollef announced that Varnish-3.0.0 is released.


1. Party Time!
--------------

I don't know if it will be too late for you now, but there will
be some partying going on at least 16 places around the world:

    http://v3party.varnish-cache.org/

This is a great chance to get to meet other local Varnishers
(or would that be Varniseurs ?) and exchange wisdom and
warstories and experience.

I'll be in Oslo, saying a couple of words and raising a pint.


2. What's new is in Varnish-3.0.0 ?
-----------------------------------

First and foremost our brand new GZIP+ESI support, this should make
it possible to serve ESI composed content while saving bandwidth
and time in usual Varnish performance style.

If you do not use ESI, it will still cut your memory usage because
Varnish will only store a gzip'ed copy of the object, and gunzip
that for the few clients which do not grok gzip (spiders mostly).

Second there are "VMODs", code-functions you can call from VCL code
to do even weirder stuff than you do with VCL.

In Varnish 1 and 2 this kind of stuff would be done with inline-C
code which, while a nice escape, is a mess by any metric.

Hopefully we will see many VMODs in the future, both from the
project and 3-rd parties.

And there is much more than that:  Real purges, half of streaming,
dynamic statistics counters (ie: per backend/health/stevedore),
VCL selectable stevedore, stronger VCL expressions and oodles
of bugfixes and much more.

Check the release notes which Tollef has faithfully assembled as I
mutilated the code.


3. Who paid for this ?
----------------------

The development of Varnish 3 has been paid for with Varnish Moral
Licenses (http://phk.freebsd.dk/VML/) and I would like to single
out the four biggest licensees, they paid for the lions part:

    Wikia, Facebook, Varnish-Software & GLOBO

(Please raise them a toast at you party!)


4. More thanks:
---------------

I also want to thank:

    The minor VML licensees (See: http://phk.freebsd.dk/VML/)
    Varnish-Software.com for hosting the project
    The Varnish-Software crew for going above and beyond
    Those who took time to report bugs and test that we fixed them.
    Everybody who helped the project along one way or another
    And finally: those who sent me thing from my Amazon wish-list
	(https://www.amazon.com/registry/wishlist/1XSZNDWXKIAS0)

Varnish may be mostly a one-man coding effort, but I could never
have done it without the rest of you.

Much appreciated.


5. VUG4
-------

The next Varnish User Group meeting will be in some kind of
symbiosis with the SURGE/2011 conference in Baltimore/USA
in the last week of september:

    http://omniti.com/surge/2011

See you there ?


6. Test please!
---------------

When I was (more) active in the FreeBSD project, I used to say
things like "Never run production on a dot-zero release." and
"Odd-numbered major releases are always trouble."

Varnish-3.0.0 is both an odd-numbered major release and a
dot-zero-dot-zero release.

But I'd like you to try to run production on it anyway, because
unless somebody does that, we will never get all the bugs found and
fixed.

It will also help you get your VCL adjusted to the changes we
have had to make in VCL syntax (this is why it is Varnish 3
instead of Varnish 2.2 btw.)


7. Next release ?
-----------------

Unless we have totally messed up, don't expect to see Varnish 4.0
until 2013, but a 3.1 or even 3.2 with new interesting features are
quite likely in 2011 and 2012.  Minor releases (3.0.1, 3.0.2...)
will happen as warranted by bugs, improvements etc.

Most of the effort right after a major release will be on fixing
bugs as they get reported, but I hope to be able to present some
kind of roadmappery for the future at the VUG4 meeting.


8. That all folks!
------------------

I hope you enjoy using Varnish 3.0.0 as much as I enjoyed writing it.

Party on!

Poul-Henning

-- 
Poul-Henning Kamp       | UNIX since Zilog Zeus 3.20
phk at FreeBSD.ORG         | TCP/IP since RFC 956
FreeBSD committer       | BSD since 4.3-tahoe
Never attribute to malice what can adequately be explained by incompetence.