Aiming for Varnish 2.1

Poul-Henning Kamp phk at
Fri Feb 12 11:49:21 CET 2010

Thanks to our new energy, we have been able to chase down and close
a fair number of weird bugs in Varnish, and right now, -trunk looks
pretty solid.

Subject to the quantum-murphy-o-meter freaking out, I think we are
ready to cut a 2.1 branch, maybe even in time for VUG2.

This will become the new "-stable" release branch, and it is probably
a good idea for you to plan to migrate to it at some point.

I really appreciate the effort some of you already are throwing at
crashing -trunk for me, and urge everybody to get in on that game
before the release.

I can't offer your cheques signed by Donald Knuth, for each bug
you find, but I really do appreciate the effort and quality of

To be honest, I have somewhat lost track of what has and what hasn't
been merged into 2.0, so I skimmed the 25k+ line diff between -trunk
and 2.0 and noticed:

 * A lot less bugs.

 * Dynamic minimum sizing of per-object storage.

 * Max number of HTTP headers a parameter.

 * -hcritbit the default, should be faster, self sizing.

 * -spersistent still aprototype, but it might be usable for a limited
   number of setups already now.

 * saint-mode.  It's complicated to explain, but smart.

 * Authentication of CLI connections.

 * New expressive ban/purge syntax

 * ban-lurker to try to keep purge-list length under control

 * New "hash" director, distributes backend traffic according to
   the object hash-key.

 * New "client" directors, distributes backend traffic according to
   client IP#.

 * Separate VCL access to request sent to backend, reply from backend.

 * Elimination of magic numbers in varnishtest testcases (+23 new tests)

 * (More?) correct handling of HTTP/1.0 <-> HTTP/1.1 headers.

 + probably a lot of details I overlooked.

 + whatever bugs we manage to fix before 2.1 is cut.

On the down-side:

 * -spersistent still not done.

 * Various solaris issues still not resolved.

 * You need to change your VCL at bit.

 * We still have a half hundred open tickets.

 * Some new bugs, didn't see them, but I know they are there.



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.

More information about the varnish-misc mailing list