WebRTC, HLS or CDN delivery: choosing a path for each participant
cRUD Research TeamThe cRUD Company
WebRTC versus HLS is often framed as a single technology choice. In a real product, it is usually a routing decision made for different participants and outputs. A presenter may need interactive, low-latency communication. A panelist may need the same. A large viewing audience may need scalable segmented delivery. A replay viewer needs a packaged on-demand asset. An external destination may impose its own ingest contract.
A CDN is not an alternative protocol in the same category. It is delivery infrastructure that can distribute HLS segments and other assets close to viewers. The useful architecture therefore asks which path each role requires, how those paths share one authoritative session and when a participant can move from one path to another without losing identity or state.
TL;DR
- WebRTC is designed for real-time media and data exchange, making it suitable for interactive participants.
- HLS transfers segmented media over ordinary web infrastructure and is commonly used for scalable live and on-demand viewing.
- A CDN distributes segments and assets; it is not a participant protocol equivalent to WebRTC or HLS.
- One session can use several delivery paths when roles, state and output identity remain coordinated.
- Choose routes by interaction, latency, scale, device, destination and continuity needs, then verify the result with evidence.
Start with the correct comparison
The W3C WebRTC Recommendation defines browser interfaces for real-time audio, video and data. HTTP Live Streaming, documented in RFC 8216, transfers a sequence of media files and playlists over HTTP. A CDN distributes those HTTP objects and other assets across a delivery network.
This means the architectural choices are related but not symmetrical. WebRTC and HLS describe different media delivery approaches. A CDN can support HLS delivery, presentation assets, recordings and replay. It can also sit beside a real-time path rather than replace it.
Use the comparison to decide the path for a participant or output, not to declare one universal winner.
Where WebRTC fits
WebRTC fits roles that must exchange media or data with low interaction delay. Presenters, panelists, moderators and small interactive groups often need this behaviour. The browser can publish and receive tracks, while the surrounding platform governs identity, permissions, subscriptions and provider operations.
Real-time participation, conversational feedback and selective media subscriptions.
The product still needs roles, session authority, reconnection, recording and operational evidence.
A large passive audience may not need the cost or state associated with an interactive path.
Recovery must restore current session state, not only create a new connection.
Where HLS fits
HLS fits live or on-demand viewing where HTTP delivery, device support and scalable distribution matter more than conversational interaction. The player obtains a playlist and requests media segments. Exact latency depends on how the stream is packaged, the player policy, the segment design, the delivery path and the live workflow.
HLS is not limited to an anonymous audience. A product can apply entitlement and signed access around it. But the media path itself does not provide the same interaction model as a real-time session. Chat, questions, polls and other application interactions normally use a separate state channel and must remain synchronized with the media timeline.
Replay and on-demand assets also fit naturally into packaged HTTP delivery, provided readiness, access and expiry are governed.
What the CDN contributes
A CDN moves cacheable segments and assets closer to viewers, absorbs repeated requests and supports distribution across locations. It can carry live HLS, replay packages, slide images, thumbnails and other objects produced by the media workflow.
The CDN does not decide who is a presenter, which participant may publish, whether a slide is authoritative, when a recording is complete or whether a degraded state is honest. Those decisions belong to the application and media platform.
A CDN is therefore an important delivery component but an incomplete orchestration layer. Treating it as the whole platform pushes session and output responsibility back into the customer product.
Choose the path by role and purpose
| Participant or output | Likely path | Primary reason |
|---|---|---|
| Presenter or panelist | WebRTC | Two-way, real-time media and control |
| Interactive attendee | WebRTC or another interactive path | Timely participation and selective subscriptions |
| Large viewing audience | HLS through a CDN | Scalable one-way program delivery |
| Presentation assets | CDN-backed objects plus state | Sharp, synchronized material without continuous video |
| Replay viewer | Packaged HTTP delivery | Controlled on-demand access and seeking |
| External destination | Destination-specific output | Fan-out to the required ingest contract |
These are common patterns, not fixed rules. Product requirements and provider support determine the approved design.
Several paths can belong to one session
A hybrid delivery architecture works only when the paths do not create separate truths. The presenter’s WebRTC session, the HLS Program Output, the slide state, the recording and the replay package should refer to the same authorised session and timeline.
This requires orchestration above the protocols. Roles and permissions determine which path a participant may use. Composition creates a governed output for large-audience delivery. State keeps application interaction aligned with media. Recording and replay retain lineage back to the live work.
The architecture can then change a delivery path without changing the meaning of the session. A viewer may move from an interactive route to scaled viewing, or recover on a lower-bandwidth route, while identity, entitlement and current state remain intact.
Plan path changes before failure
Path selection should not happen only at the beginning. Devices change orientation, networks become unreliable, audiences grow and participants change roles. The platform needs policies for moving between approved states.
- Define which transitions are allowed for each role.
- Preserve the authoritative session and current program position.
- Make degraded states visible rather than presenting stale media as live.
- Keep interaction state separate enough to survive a media-path change.
- Record why the route changed and what the participant received.
- Do not promise seamless switching unless the specific workflow has been tested.
Continuity is a designed outcome, not an automatic property of a protocol.
How cRUD Morph governs delivery
Deliver is the cRUD Morph capability that selects and operates approved paths for participants, program outputs and destinations. It works with Orchestrate for session authority, Adapt for current conditions, Compose for Program Output, Continuity for disruption and Assure for operational evidence.
This model allows WebRTC, streaming, CDN distribution and external destinations to serve different purposes inside one media lifecycle. It does not claim that every path, provider or transition is available in every engagement. Those details are confirmed during technical discovery and tested for the approved environment.
Summary
WebRTC, HLS and a CDN do different jobs. WebRTC supports real-time participation. HLS supports segmented live and on-demand delivery. A CDN distributes media and assets efficiently. A product may need all three.
The architectural goal is not to force one route onto everyone. It is to choose an approved path for each role and output while keeping one authoritative session, one understandable timeline and evidence of what each participant received.
Frequently asked questions
What is the main difference between WebRTC and HLS?
Is a CDN an alternative to WebRTC?
Can one live session use both WebRTC and HLS?
Which delivery path has lower latency?
How does cRUD Morph choose a delivery path?
About this article. This article explains common protocol and delivery roles. It does not provide a universal performance comparison or guarantee a latency, scale or availability outcome. Results depend on implementation, provider, location, device, network and workflow.