Talk Selection & CFP Review

Overview

How different conferences select and schedule talks.

Selection models

FOSDEM (65 tracks, ~1000 talks):

• Staff picks ~40 main talks
• 65 devrooms each have own CFP and managers
• Devrooms can recommend talks for main stage

COSCUP (23 tracks, 30 rooms, 340 talks):

• Invite community review teams per track
• Staff makes final decision

OpenFest:

• Program committee of outside community members
• Vote after CFP closes to catch blind spots
• 45-minute talks, 2-4 hour workshops

DevConf.IN:

• 5-6 tracks
• Require 40% new speakers
• Anonymous first-pass review

Chemnitzer Linux-Tage:

• No fixed tracks
• Talks are 30, 60, or 90 minutes
• Custom system for everything

Talk formats

• 20 minutes: Intro talks and pitches
• 40 minutes: Standard length
• Lightning talks: 5 or 15 minutes
• Workshops: 1 to 4 hours

Lightning talk selection methods:

• Chemnitzer Linux-Tage: Postcards, random selection
• OpenFest: First come, first serve sign-up
• FOSSASIA: Pre-record, publish automatically
• OpenChain: Attendee voting

Keynotes and invited talks are staff picks, sometimes intentional to fill gaps or balance representation.

Common challenges

• Abstract quality: Not all good speakers write good abstracts
• Recycled talks: Some events limit repeats to 30%. Others ask for recordings.
• Topic trends: Everyone wants Kubernetes and DevOps. Harder to find other topics.
• Celebrity speakers: Some draw crowds regardless of topic quality
• New vs returning: DevConf.IN targets 40% new speakers. Hard to balance. Wanted 50%, but was impossible
• Prolific submitters: Some people submit dozens of proposals (DevConf had one with 14)
• AI-generated submissions: Starting to appear. Some events detect or flag.
• Visa requirements: Need advance notice and invitation letters

Review process

• Anonymous review - At least for first pass
• Community input - Voting or public review catches blind spots
• Check history - Look at previous talks and publications
• Scale reviewers - DevConf.IN uses 1 reviewer per 10 talks

Scheduling

• Speaker availability: FOSDEM asks, many don’t bother
• Slot selection: DevConf does first-come first-served
• Track conflicts: Hard to avoid with parallel sessions

What works

• Clear criteria applied consistently
• Being flexible about what’s interesting (not just trendy)
• Help speakers with visas, abstract writing, community intro
• Pick diverse speakers, not just people you know
• Multiple reviewers instead of one person deciding