Return to Blog
September 25, 2014
3 min read time

Varnish Developer News: VDD14Q3 @ VG.no/Schibsted in Oslo, Norway

Rubén Romero
  • Varnish Cache
  • development

    • Once every quarter there is a VDD - Varnish Developer Day. We typically gather between 8-18 people that are interested in Varnish development and discuss Varnish development. unless you want to use your day in Varnish packaging, internals or VCL design, you are better off just reading this post and/or the notes from the meeting :-)

    The last VDD was held in Oslo, in the offices of VG.no and Schibsted at Akersgata. After that we went to aDevOps meetup and finished the evening going for some food and drinks.

    You can see the notes as they were taking throughout the day on the VDD14Q3 page. 

    The main discussion points were (in chronological order):

    • Next meeting, VDD14Q4 is scheduled for the week of November 20. This wil happen together with VUG9 in Barcelona, Spain following the Velocity Conference on November 17-19.
    • Backend connection request timeouts. You can read the background of the discussion in the list.
      • Decision: Close more intelligently - ditch rest of list whenever we get a new connection.
      • A fullpackage of backend connection handling to be discussed as well.
    • Stale-While-Revalidate in Varnish. Background is also available on the list.
      • Decision: Put it in Varnish after reviewing the latest patch-
      • Consider putting stale-if-error later.
    • Saint Mode in Varnish 4.0. Background available here.
      • Decision: Add the hash key to the busyobj, so that Saint Mode can be implemented on a VMOD.
    • VMOD Packaging
      • Decision: Agree on a common package name for DEB and RPM distributions.
      • Add the PKG_CONFIG_PATH patch to the documentation.
    • Adding VMOD functionality to VMOD Standard and the Varnish Cache disttribution
      • Decision1: Add cachereqbody to vmod-std 
      • Decision2: Certain VMODs will be pulled into Varnish Cache. For now we start with vmod-cookie after teaching it about Set-Cookie hearders.
    • VMOD Miscellaneous
      • Decision1: Agree on a common package name for DEB and RPM distributions vmod-<name>
      • Decision2: PRIV_REQ memory storage is needed for vmod-cookieand need to be implemented in Varnish. PHK will do it unless other do it before him.
    • 4.0 Merging.
    • Varnish 4.1 Planning
    • Surrogate Control. Sections 2.2 and 2.3 of the Edge Architecture Specification.
      • Decision: We'll do this, and make it a configurable in 4.0.x/4.1. Default on in (future) 5.0. If surrogate-control is there, prefer that instead of cache-control.
      • Standarization of Surrogate-Control Header and Targetting in builtin.vcl (what we called default.vcl in Varnish 3.0 and older).

    Besides working through these points, we did a Mini Bug Wash as well as discussed PCRE limitations, Phantom 503s, Header Token editing and other subjects upon which a decision was not made.