Case 761
The rotation covered the services in the registry
the_rotation_covered_the_services_in_the_registry.eml - Every service in the registry has a named on-call rotation with a tested escalation path, and no page has gone unanswered in fourteen months. Which services are in the registry is computed below.
ok: true — round-trip fixpoint reached (python1 == python2)updated 2026-09-08
EML
eml# Self-authored for the EML case corpus (no external origin). Every service in
# the registry has a named on-call rotation with a tested escalation path, and
# no page has gone unanswered in fourteen months. Which services are in the
# registry is computed below.
#
# The rotation is properly run. Every entry names a person rather than a team
# alias that resolves to nobody at three in the morning; the escalation path is
# tested monthly by paging it for real; the handover is a scheduled event with
# a checklist rather than a calendar colour; and in fourteen months no page has
# gone unanswered and no escalation has run out of levels.
#
# A service is in the rotation because it is in the registry, and it is in the
# registry because somebody added it. A service nobody registered does not have
# a gap in its coverage; it has no row.
#
# The deployment platform is running one hundred and forty services.
96 => services_in_the_registry
96 => services_with_a_named_rotation
140 => services_running_on_the_platform
14 => months_without_an_unanswered_page
12 => escalation_tests_last_year
0 => alerts_when_a_running_service_is_absent_from_the_registry
0 => jobs_that_enumerate_the_platform
1 => registrations_that_are_a_manual_step
services_running_on_the_platform - services_in_the_registry => services_running_without_a_rotation
int(services_in_the_registry * 10000 / services_running_on_the_platform) => covered_per_myriad
int(services_with_a_named_rotation * 10000 / services_in_the_registry) => registry_coverage_per_myriad
"services in the registry : " + str(services_in_the_registry) ^0
" with a named rotation : " + str(services_with_a_named_rotation) ^0
" registry coverage : " + str(registry_coverage_per_myriad) + " per ten thousand" ^0
"months without an unanswered page : " + str(months_without_an_unanswered_page) ^0
"escalation tests last year : " + str(escalation_tests_last_year) ^0
"" ^0
"services running on the platform: " + str(services_running_on_the_platform) ^0
" covered by a rotation : " + str(services_in_the_registry) ^0
" running without one : " + str(services_running_without_a_rotation) ^0
" share covered : " + str(covered_per_myriad) + " per ten thousand" ^0
"" ^0
"registration is a manual step : " + str(registrations_that_are_a_manual_step) ^0
"jobs that enumerate the platform: " + str(jobs_that_enumerate_the_platform) ^0
"alerts when a running service is unregistered : " ^0
" " + str(alerts_when_a_running_service_is_absent_from_the_registry) ^0
"" ^0
# ---- what the rotation verified ----
"the on-call rotation" ^0
" each entry names : a person, not a team alias that" ^0
" resolves to nobody at three in the morning" ^0
" escalation path : tested by paging it for real," ^0
" " + str(escalation_tests_last_year) + " times last year" ^0
" handover : a scheduled event with a checklist" ^0
" months without an unanswered page : " + str(months_without_an_unanswered_page) ^0
" escalations that ran out of levels : none" ^0
" verdict : ANSWERED" ^0
"" ^0
" paging the escalation path for real, monthly, is the" ^0
" part almost nobody does, and it is why the fourteen" ^0
" months mean something" ^0
"" ^0
# ---- how a service gets a rotation ----
"membership" ^0
" a service has a rotation because : it is in the" ^0
" registry" ^0
" it is in the registry because : somebody added it," ^0
" " + str(registrations_that_are_a_manual_step) + " manual step" ^0
" a registered service with a broken rotation : visible," ^0
" and the monthly test finds it" ^0
" an unregistered service : has no row to be broken" ^0
" services in that state : " + str(services_running_without_a_rotation) ^0
"" ^0
" the coverage figure is complete over the registry and" ^0
" the registry is complete over what people remembered" ^0
"" ^0
# ---- the coverage number is a tautology ----
"the reported coverage" ^0
" numerator : services with a named rotation" ^0
" denominator : services in the registry" ^0
" what adding a service to the registry requires : giving" ^0
" it a rotation" ^0
" so the ratio : " + str(registry_coverage_per_myriad) + " per ten thousand, necessarily" ^0
" could it ever be lower : only if a registered service" ^0
" were left without one, which the form prevents" ^0
" what it therefore measures : nothing that can vary" ^0
"" ^0
# ---- what a page from the forty-four looks like ----
"an alert from an unregistered service" ^0
" does the service emit alerts : yes, the platform" ^0
" scrapes it like any other" ^0
" where does the alert route : the default receiver" ^0
" who reads the default receiver : it is a channel, not a" ^0
" person" ^0
" does it page : no" ^0
" does it count against the unanswered-page record : no;" ^0
" it was never a page" ^0
"" ^0
# ---- null control ----
# The same rotation, with a job that enumerates the platform's own service list
# and raises anything running without a registry entry.
services_running_on_the_platform => nc_services_the_check_can_see
0 => nc_services_running_without_a_rotation
1 => nc_jobs_that_enumerate_the_platform
"null control - enumerate the platform, diff the registry" ^0
" escalation tests : " + str(escalation_tests_last_year) + ", unchanged" ^0
" jobs that enumerate the platform : " + str(nc_jobs_that_enumerate_the_platform) ^0
" services the check can see : " + str(nc_services_the_check_can_see) ^0
" running without a rotation : " + str(nc_services_running_without_a_rotation) ^0
" the rotation did not improve; the population it is" ^0
" measured against stopped being the list it is drawn" ^0
" from" ^0
"" ^0
# ---- the rule ----
"what a fully covered rotation guarantees" ^0
" every registered service has someone who answers :" ^0
" exactly, tested monthly, " + str(months_without_an_unanswered_page) + " months clean" ^0
" every service has someone who answers : not addressed;" ^0
" the roster is built from a list people maintain, and" ^0
" a service that is not on it is not uncovered" ^0
"" ^0
"a coverage ratio whose denominator is the same list as its" ^0
"numerator cannot fall; the number that can is a comparison" ^0
"against something the registry did not write, and nothing" ^0
"here forms it" ^0
"" ^0
"Every entry names a person rather than an alias, the escalation path is paged" ^0
"for real " + str(escalation_tests_last_year) + " times a year, handover has a checklist, and " + str(months_without_an_unanswered_page) + " months have passed" ^0
"with no unanswered page. A service is on the roster because it is in a registry" ^0
"somebody maintains by hand, so coverage reads " + str(registry_coverage_per_myriad) + " per ten thousand of " + str(services_in_the_registry) ^0
"services while " + str(services_running_on_the_platform) + " run - " + str(services_running_without_a_rotation) + " of them under " + str(alerts_when_a_running_service_is_absent_from_the_registry) + " alerts." ^0Python (deterministic transpilation)
pythonservices_in_the_registry = 96
services_with_a_named_rotation = 96
services_running_on_the_platform = 140
months_without_an_unanswered_page = 14
escalation_tests_last_year = 12
alerts_when_a_running_service_is_absent_from_the_registry = 0
jobs_that_enumerate_the_platform = 0
registrations_that_are_a_manual_step = 1
services_running_without_a_rotation = services_running_on_the_platform - services_in_the_registry
covered_per_myriad = int(services_in_the_registry * 10000 / services_running_on_the_platform)
registry_coverage_per_myriad = int(services_with_a_named_rotation * 10000 / services_in_the_registry)
print("services in the registry : " + str(services_in_the_registry))
print(" with a named rotation : " + str(services_with_a_named_rotation))
print(" registry coverage : " + str(registry_coverage_per_myriad) + " per ten thousand")
print("months without an unanswered page : " + str(months_without_an_unanswered_page))
print("escalation tests last year : " + str(escalation_tests_last_year))
print("")
print("services running on the platform: " + str(services_running_on_the_platform))
print(" covered by a rotation : " + str(services_in_the_registry))
print(" running without one : " + str(services_running_without_a_rotation))
print(" share covered : " + str(covered_per_myriad) + " per ten thousand")
print("")
print("registration is a manual step : " + str(registrations_that_are_a_manual_step))
print("jobs that enumerate the platform: " + str(jobs_that_enumerate_the_platform))
print("alerts when a running service is unregistered : ")
print(" " + str(alerts_when_a_running_service_is_absent_from_the_registry))
print("")
print("the on-call rotation")
print(" each entry names : a person, not a team alias that")
print(" resolves to nobody at three in the morning")
print(" escalation path : tested by paging it for real,")
print(" " + str(escalation_tests_last_year) + " times last year")
print(" handover : a scheduled event with a checklist")
print(" months without an unanswered page : " + str(months_without_an_unanswered_page))
print(" escalations that ran out of levels : none")
print(" verdict : ANSWERED")
print("")
print(" paging the escalation path for real, monthly, is the")
print(" part almost nobody does, and it is why the fourteen")
print(" months mean something")
print("")
print("membership")
print(" a service has a rotation because : it is in the")
print(" registry")
print(" it is in the registry because : somebody added it,")
print(" " + str(registrations_that_are_a_manual_step) + " manual step")
print(" a registered service with a broken rotation : visible,")
print(" and the monthly test finds it")
print(" an unregistered service : has no row to be broken")
print(" services in that state : " + str(services_running_without_a_rotation))
print("")
print(" the coverage figure is complete over the registry and")
print(" the registry is complete over what people remembered")
print("")
print("the reported coverage")
print(" numerator : services with a named rotation")
print(" denominator : services in the registry")
print(" what adding a service to the registry requires : giving")
print(" it a rotation")
print(" so the ratio : " + str(registry_coverage_per_myriad) + " per ten thousand, necessarily")
print(" could it ever be lower : only if a registered service")
print(" were left without one, which the form prevents")
print(" what it therefore measures : nothing that can vary")
print("")
print("an alert from an unregistered service")
print(" does the service emit alerts : yes, the platform")
print(" scrapes it like any other")
print(" where does the alert route : the default receiver")
print(" who reads the default receiver : it is a channel, not a")
print(" person")
print(" does it page : no")
print(" does it count against the unanswered-page record : no;")
print(" it was never a page")
print("")
nc_services_the_check_can_see = services_running_on_the_platform
nc_services_running_without_a_rotation = 0
nc_jobs_that_enumerate_the_platform = 1
print("null control - enumerate the platform, diff the registry")
print(" escalation tests : " + str(escalation_tests_last_year) + ", unchanged")
print(" jobs that enumerate the platform : " + str(nc_jobs_that_enumerate_the_platform))
print(" services the check can see : " + str(nc_services_the_check_can_see))
print(" running without a rotation : " + str(nc_services_running_without_a_rotation))
print(" the rotation did not improve; the population it is")
print(" measured against stopped being the list it is drawn")
print(" from")
print("")
print("what a fully covered rotation guarantees")
print(" every registered service has someone who answers :")
print(" exactly, tested monthly, " + str(months_without_an_unanswered_page) + " months clean")
print(" every service has someone who answers : not addressed;")
print(" the roster is built from a list people maintain, and")
print(" a service that is not on it is not uncovered")
print("")
print("a coverage ratio whose denominator is the same list as its")
print("numerator cannot fall; the number that can is a comparison")
print("against something the registry did not write, and nothing")
print("here forms it")
print("")
print("Every entry names a person rather than an alias, the escalation path is paged")
print("for real " + str(escalation_tests_last_year) + " times a year, handover has a checklist, and " + str(months_without_an_unanswered_page) + " months have passed")
print("with no unanswered page. A service is on the roster because it is in a registry")
print("somebody maintains by hand, so coverage reads " + str(registry_coverage_per_myriad) + " per ten thousand of " + str(services_in_the_registry))
print("services while " + str(services_running_on_the_platform) + " run - " + str(services_running_without_a_rotation) + " of them under " + str(alerts_when_a_running_service_is_absent_from_the_registry) + " alerts.")stdout (executed)
textservices in the registry : 96
with a named rotation : 96
registry coverage : 10000 per ten thousand
months without an unanswered page : 14
escalation tests last year : 12
services running on the platform: 140
covered by a rotation : 96
running without one : 44
share covered : 6857 per ten thousand
registration is a manual step : 1
jobs that enumerate the platform: 0
alerts when a running service is unregistered :
0
the on-call rotation
each entry names : a person, not a team alias that
resolves to nobody at three in the morning
escalation path : tested by paging it for real,
12 times last year
handover : a scheduled event with a checklist
months without an unanswered page : 14
escalations that ran out of levels : none
verdict : ANSWERED
paging the escalation path for real, monthly, is the
part almost nobody does, and it is why the fourteen
months mean something
membership
a service has a rotation because : it is in the
registry
it is in the registry because : somebody added it,
1 manual step
a registered service with a broken rotation : visible,
and the monthly test finds it
an unregistered service : has no row to be broken
services in that state : 44
the coverage figure is complete over the registry and
the registry is complete over what people remembered
the reported coverage
numerator : services with a named rotation
denominator : services in the registry
what adding a service to the registry requires : giving
it a rotation
so the ratio : 10000 per ten thousand, necessarily
could it ever be lower : only if a registered service
were left without one, which the form prevents
what it therefore measures : nothing that can vary
an alert from an unregistered service
does the service emit alerts : yes, the platform
scrapes it like any other
where does the alert route : the default receiver
who reads the default receiver : it is a channel, not a
person
does it page : no
does it count against the unanswered-page record : no;
it was never a page
null control - enumerate the platform, diff the registry
escalation tests : 12, unchanged
jobs that enumerate the platform : 1
services the check can see : 140
running without a rotation : 0
the rotation did not improve; the population it is
measured against stopped being the list it is drawn
from
what a fully covered rotation guarantees
every registered service has someone who answers :
exactly, tested monthly, 14 months clean
every service has someone who answers : not addressed;
the roster is built from a list people maintain, and
a service that is not on it is not uncovered
a coverage ratio whose denominator is the same list as its
numerator cannot fall; the number that can is a comparison
against something the registry did not write, and nothing
here forms it
Every entry names a person rather than an alias, the escalation path is paged
for real 12 times a year, handover has a checklist, and 14 months have passed
with no unanswered page. A service is on the roster because it is in a registry
somebody maintains by hand, so coverage reads 10000 per ten thousand of 96
services while 140 run - 44 of them under 0 alerts.Trace event types
eml:run:starteml:assigneml:outputeml:run:done