“RSC or SSR?”
RSC is an execution/payload boundary; SSR produces initial HTML. A framework can use both.
Quick reference · Print-friendly
Rendering patterns field guide
Choose per route or region. Start from constraints; combine patterns deliberately.
Comparison matrix
| Pattern | HTML timing | Primary gain | Primary cost | Good signal |
|---|---|---|---|---|
| CSR | Browser runtime | App-like client model | Startup JS and data waterfall | Authenticated tool; repeat use |
| SSR | Each request | Fresh or personalized HTML | Compute, latency, operations | Cookies or request data shape output |
| SSG | Build time | Cheap, resilient CDN delivery | Staleness and build scale | Same content for many users |
| ISR | Build + regeneration | Static reads with bounded freshness | Invalidation and stale windows | Many pages; updates tolerate delay |
| Streaming SSR | Request, chunked | Useful shell before slow regions | Fallback and infra complexity | Independent regions have uneven latency |
| Islands | Usually build/server | JavaScript only for widgets | Cross-island coordination | Mostly content, sparse interaction |
| Server Components | Server execution | Server-only dependencies and data access | Boundary mental model and framework coupling | Large non-interactive component subtrees |
| Resumability | Server + lazy continuation | Avoid eager hydration replay | Serialization constraints and ecosystem | Startup JS is the bottleneck |
Interview decision script
- Classify the content: public or private, shared or personalized, stable or fresh?
- Find the interaction boundary: what truly requires browser state, events, or browser APIs?
- Choose HTML timing: build, request, or browser—and at which route/component boundary?
- Choose reuse: CDN, application cache, per-user response, and explicit invalidation semantics.
- Choose delivery: complete response or streamed boundaries with meaningful fallbacks?
- Budget activation: no JS, progressive enhancement, islands, hydration, or resumability?
- Name failure modes: stale data, hydration mismatch, buffering, cache leakage, server overload, JS disabled.
- Name evidence: TTFB, LCP, INP, JS bytes, cache-hit rate, origin load, and conversion/task success.
Correct the premise
“Static or dynamic?”
Usually a spectrum by region and data source, not a whole-site binary.
“SSR fixes SEO?”
It can expose initial content, but correctness, crawlability, metadata, links, status codes, and speed still matter.
“Streaming makes it faster?”
It can improve progressive display; it does not reduce all work and may be neutralized by buffering.