httpuv
httpuv ships almost nothing, which is what you want from the web server under Shiny
Comparison · DevOps
callr vs later
A side-by-side editorial comparison of callr and later — release velocity, themes, recent moves, and the top alternatives to consider.
callr vs later: at a glance
What is callr?
callr adds OpenTelemetry tracing, making R subprocess work observable across process boundaries
callr runs R code in fresh subprocesses and had been in maintenance for roughly two years — the 3.7.x releases are error-object polish, DLL handling, and Windows fixes. 3.8.0 breaks that pattern with a batch of substantive changes: pseudo-terminal support, carrier::crate() handling, configurable temp directories, honest non-zero exit statuses, and OpenTelemetry instrumentation.
What is later?
later spends its releases keeping R's event loop compiling against moving toolchains.
later provides the event loop that lets R schedule callbacks, underpinning shiny, httpuv and most async work in the language. Recent releases are overwhelmingly build and compatibility work: Rcpp error-handling changes, a C23 namespace collision with glibc, and header ordering to satisfy Rcpp auto-includes.
callr vs later: editorial side-by-side
callr
0.0DEVOPS
callr adds OpenTelemetry tracing, making R subprocess work observable across process boundaries
◆ Current state
callr runs R code in fresh subprocesses and had been in maintenance for roughly two years — the 3.7.x releases are error-object polish, DLL handling, and Windows fixes. 3.8.0 breaks that pattern with a batch of substantive changes: pseudo-terminal support, carrier::crate() handling, configurable temp directories, honest non-zero exit statuses, and OpenTelemetry instrumentation.
◆ Where it's heading
The package is being repositioned from a convenience wrapper into infrastructure that production R workloads can be operated against. Tracing spans that propagate a W3C traceparent into the child process, plus exit codes that finally reflect failure, are both about making callr legible to systems outside R. The pty and libpath changes point the same way — toward callr hosting long-lived, externally managed sessions.
◆ Prediction
Expect the OpenTelemetry surface to spread to sibling packages in the processx/r-lib process stack, and further work on subprocess failure semantics now that non-zero exits are the default.
later
0.0DEVOPS
later spends its releases keeping R's event loop compiling against moving toolchains.
◆ Current state
later provides the event loop that lets R schedule callbacks, underpinning shiny, httpuv and most async work in the language. Recent releases are overwhelmingly build and compatibility work: Rcpp error-handling changes, a C23 namespace collision with glibc, and header ordering to satisfy Rcpp auto-includes.
◆ Where it's heading
The API is settled enough that the release calendar is set by the surrounding ecosystem rather than by any feature plan. 1.4.5 cleaned house by dropping pre-unwind-protect fallbacks, and 1.4.6 carries the only recent change a user would notice at all — lower CPU use when idle at the console.
◆ Prediction
Expect further releases pinned to compiler and Rcpp changes, with the observable surface staying essentially where it is.
Alternatives to callr and later
Other DevOps products tracked by Sparkpulse, ranked by recent ship velocity. Each card links to a full editorial trajectory and lets you pivot into a head-to-head comparison with either callr or later.
promises
promises drops R6 for plain environments and wires async work into OpenTelemetry
withr
withr rebuilt defer() on base R's own on.exit(), and got much faster for it
htmltools
htmltools spends three releases walking back its own fill-layout opinions
xml2
xml2 relicensed to MIT, then moved its hot accessors from S3 into C
httr
httr's last releases are housekeeping - the package now points users at httr2
Recent activity from callr and later
Latest ship moves from both products, interleaved chronologically. ⚡ = editorial spark.
- 2mo agocallrOpenTelemetry tracing, pty support, and honest subprocess exit codes ⚡
- 5mo agolaterlater 1.4.8 tracks Rcpp Rf_error handling changes
- 5mo agolaterlater 1.4.7 fixes a C23 once_flag collision breaking builds
- 6mo agolaterlater 1.4.6 improves idle responsiveness at the R console
- 7mo agolaterlater 1.4.5 drops legacy fallbacks, requires R 3.5 and Rcpp 1.0.10
- 11mo agolaterlater 1.4.4 fixes test timings only
- 11mo agolaterlater 1.4.3 stops touching .Random.seed, drops autorun argument
- 2y agocallrEnv var to skip copying the client DLL
- 2y agocallrRunning-time fix and fully qualified subprocess calls
- 3y agocallrErrors carry subprocess stdout and stderr again
- 3y agocallrExperimental add_hook() for subprocess startup
- 4y agocallrUncompressed transport by default; revamped error objects
Frequently asked questions
What is the difference between callr and later?
They serve adjacent needs but don't currently overlap on shipped themes. callr and later are shipping at a similar cadence (velocity 0.0 vs 0.0, both within Sparkpulse's "active" band). See the at-a-glance table above for a side-by-side breakdown of velocity, recent sparks, and editorial themes.
Is callr better than later?
Sparkpulse doesn't pick a winner — we score release velocity, not feature parity. callr and later are shipping at a similar cadence (velocity 0.0 vs 0.0, both within Sparkpulse's "active" band). For your specific use case, the alternatives sections above list other DevOps products to evaluate alongside.
What are the best alternatives to callr?
Top callr alternatives in DevOps are ranked by recent ship velocity. Browse the "callr alternatives" section above for the current picks, or visit /alternatives/callr for the full list with editorial commentary on each.
What are the best alternatives to later?
Top later alternatives in DevOps are ranked by recent ship velocity. Browse the "later alternatives" section above for the current picks, or visit /alternatives/later-r for the full list with editorial commentary on each.